{ "text": "So , Deepseek version 4 Pro is officially out today .\nNow , you might be confused because this model was technically\nout , but it was in general availability , but this is the official\nrelease , meaning that they have trained the model a little bit\nmore and produced a stronger version of the model that is available\ntoday .\nNow , if I were to summarize today's release in one simple sentence\n, it is that the price to performance is becoming a very ,\nvery important thing because not only do we have a new model from\nthe Deepseek team , the SpaceX team also dropped Grock 4 .\n6 six .\nAnd both of these models are competing with the best Frontier Labs\nat a fraction of their cost .\nLet's start with Gro 4 .\n6 .\nAnd one thing I'm going to say is that I'm genuinely surprised\nby the SpaceX team .\nI'm not trying to glaze them or Elon Musk or anything .\nI was just very critical of this lab .\nIn 2025 , they dropped Grog 4 and other models like that ,\nbut I wasn't really , you know , mind blown with their performance\nbecause they're pretty they're pretty subpar compared to any of\nthe other labs out there .\nBut in 2026 , it looks like things have kind of changed a little\nbecause Grock 4 .\n5 was quite competitive based off what it cost and Grock 4 .\n6 is actually not too bad .\nIf we take a look at the benchmarks , for example ,\nif we start with the artificial analysis intelligence index ,\nthis model achieves a 61 and Fable 5 is at 62 .\nNow , what's really important to remember is that once again ,\nthis model is quite cheap compared to Fable 5 .\nThis is about 2 per million input tokens and 6 per million output\ntokens .\nIn Fable 5 sits at 10 per million input tokens and 50 per million\noutput tokens .\nAnd this is why you start to appreciate this release a little bit\nmore .\nIt might not beat the performance of the best models is matching\nthem .\nEven GPT 5 .\n6 so which is a pretty capable model and this is set at max on\nthe artificial analysis index .\nThis model achieves 61 and Grok 4 .\n6 61 .\nSo it ties it and it's much cheaper .\nAnd then even on all of these other benchmarks ,\nfor example , the Code Val one , it actually beats Fable 5 ,\nwhich is at 1741 .\nAnd then GPT 5 .\n6 , it's at 1728 .\nI'm not sure why they didn't choose Opus 5 as well ,\nbut I guess they wanted to choose the quote unquote strongest model\nlineup from each lab .\nAnd they chose Fable 5 for Enthropic , which is fair .\nAnd then Deep Software Engineering one , which is a critical benchmark\n.\nThis model doesn't beat Fable 5 or GPT 5 .\n6 six soul , but it gets close to it .\nIt's 65 .\n9 and Fable 5 sits at 70 .\nBut if you're getting results that are pretty close and the model\nis five times cheaper , I wouldn't be too disappointed with this\nresult .\nAnd then same thing with Cursor Bench 3 .\n2 , the model achieves 69 .\n9 , funny number .\nAnd Fable 5 sits at 70 .\n5 .\nSo once again , closer .\nAnd Grock 4 .\n6 beats GPT 5 .\n6 .\nSame thing with the Frontier Code .\nIt gets close to Fable 5 , beats GPT 5 .\n6 .\nSo yeah , this model is actually available in cursor .\nSo the partnership with cursor or I guess the acquisition has really\nhelped SpaceX make some strides in the AI space this year and Grock\nbuild is something that you know maybe not a lot of us have been\nusing so far but it's probably going to be another platform like\ncodeex or cloud code that we start to use but obviously cursor\nis quite strong as well .\nSo you have options available for you to use them in both .\nAnd one thing to note is that they're offering two times usage\ninside Grok Build and Cursor for the first week .\nSo if you just want to try it out , see what you feel about it\n, then you know it might be worth trying it out right now cuz you\nget double the usage in the first week .\nNow if we were to take a look at some of the outputs that people\nhave been generating with Grok 4 .\n6 , what we're looking at right now is a Falcon 9 booster return\nsequence simulation .\nAnd this was done in a single HTML file .\nAnd as I said , if you expected to get this type of output from\nGrock in 2025 , you would be kind of surprised because you wouldn't\nexpect something like this to be generated with Grock .\nBut now it looks like we have to start taking the Grok team a\nlittle bit more serious because this output is quite competitive\n.\nAnd obviously , we're just looking at a simulation and we're just\nbasing it off of a visual representation .\nBut if the model is able to produce something like this consistently\n, then I would expect a lot of people to start adopting Grock because\nnumber one , it is cheaper than the other labs at least at the\nmoment because we don't know if this pricing strategy is going\nto be sustainable for the SpaceX team in the long run .\nBut at least for now , their models are definitely cheaper compared\nto the others .\nAnd as I mentioned , this model excelled at the artificial analysis\nindex .\nThis model jumped to probably number four model .\nIt's tied to GPT 5 .\n6 pretty much similar .\nSo you could say number three as well , but the models before that\nare Fable 5 and Opus 5 , which are only above the model by about\na 1 or a 2 difference .\nSo yeah , even on this intelligence index , which if you're not\nfamiliar with has nine evaluations .\nSo on all of these evaluations , it's kind of matching almost Fable\n5 performance , which is crazy to see .\nBefore we continue , we just launched the Universe of AI newsletter\n.\nIf you want to stay on top of AI news without having to hunt for\nit , link is in the description .\nDon't miss out .\nAnd what you see on screen right now is a racing game that Grok\n4 .\n6 build .\nAnd based off of this post , the model took about 1 minute and\nit was a fiveword prompt , which was a create a simple racing game\nin HTML .\nSo , if you're able to generate something like this easily using\nGrock 4 .\n6 , I think a lot of people will be happy .\nAnd this is a more detailed analysis of what it costs to run GPT\n5 .\n6 6 on the artificial analysis index and what it produced meaning\nthe output .\nBoth of these models if you remember scored 61 on the artificial\nanalysis index .\nNow to run the whole test with GPT 5 .\n6 it cost about 2 .\n6 it cost about 2 .\n6 it cost about 1 .\n1Kish .\nAnd this tells you that you're getting similar level of performance\nat half the cost .\nSo yeah , this is a big release for the SpaceX team because they\njust proved once again that they are a lab that you seriously start\nneed to considering , especially in 2026 .\nAnd I'm going to talk more about Deep Seek version 4 Pro GA .\nBut basically what we're seeing today is that both of these releases\nkind of emphasize the fact that performance and all above that\nis the price at what you're getting for that performance is becoming\nmore and more important for all users because we see many labs\nnow focusing on creating the best model at the cheapest cost .\nLast year in 2025 most of the labs were just focused on I would\nsay creating the strongest model .\nYes , cost was important , but I think most of the times the frontier\nlabs , meaning OpenAI , Anthropic , were kind of more lenient on\nthat fact because they didn't have as strong of a competition .\nIntelligent models that are maybe not always ahead of OpenAI Enthropic\n, but match their performance at a fraction of the cost .\nSo yes , price to performance ratio is becoming a critical I would\nsay indicator in 2026 .\nNow , this is the updated benchmark chart after the release of\nthe new model .\nAnd one thing you'll see across the board is that it matches the\ntop level performance of many of the models .\nThe one thing interesting over here is that they haven't put Opus\n5 here for some reason .\nThere is Fable 5 here that we can compare this model against .\nBut one thing you'll notice is that DeepSeek ,\nremember this model costs 43 .\n5 cents per million input tokens and 87 cents per million output\ntokens .\nWhile the other models all over here are way more expensive than\nthat .\nSo the first thing if you look at terminal bench 2 .\n1 the model scores 87 .\n9 .\nThe older version of the model was 72 .\n1 and the flash version which we got last week was 82 .\n7 .\nAnd what's crazy is that Fable 5 is 88 .\nYes , 88 .\nSo this model is only .\n1 behind Fable 5 .\nAnd then on the Cyber Gym , which is Cyber Security ,\nthe model actually beats Fable 5 .\nFable 5 sits at 83 .\n1 while Deepseek version 4 Pro , the new one that we got today\n3 .\n3 .\nNow , if this tells you something is that this model is once again\nreally geared at cyber security .\nWe saw the Flash model also be geared towards cyber security and\nbecoming a model that was quite capable in that area .\nAnd we're seeing the same thing today with the new version 4 Pro\nwhich sits at 83 .\n3 outperforming Fable 5 on the benchmark .\nOn deep software engineering , this model is not beating Fable\n5 because Fable 5 sits at 70 , but the model achieves 62 .\n7 which is ahead of GLM 5 .\n2 and a bit behind Kimmy K3 which sits at 67 .\n5 .\nBut one thing again , this is a big jump compared to the preview\nversion of the model which was at 12 .\n8 .\nSo yeah , they have really trained this model and they have improved\nit on the back end because we can clearly see in this deep software\nengineering benchmark .\nAnd there's a couple of other benchmarks , but one key benchmark\nthat this model excels at is the automation bench where the model\nachieves 31 .\n8 and Fable 5 29 .\n1 .\nSo once again , the model outperforms Fable 5 .\nNow this is a fraction of the cost as I mentioned .\nThis is 43 cents and 87 for input and output respectively versus\nFable 5 which is at 10 .\n50 .\nSo yes , if we look at the terminal bench for example ,\nwe are seeing this model achieve a result which is 0 .\n1 behind the best model out there Fable 5 and the pricing is insane\nto look at 43 versus 10 87 versus 50 per million input tokens .\nSo when we turn that into a fraction , this model is 57 times cheaper\n.\nAnd earlier last week , I think I made a video about how DeepSeek\nversion 4 Pro was going to achieve a result like this because there\nwas a investor report that leaked .\nAnd when I saw the pricing where it said that it was going to match\nFable 5 or even outperform it and be at 57 times cheaper per cost\n.\nWhen I was reading the investor report , I didn't really believe\nit .\nBut today , it is proven because at least on the benchmark so far\n, yes , I'm just saying the benchmarks , we are seeing similar\nlevel performance .\nNow , if this model actually performs like that in production ,\nit's a little too early to tell yet .\nWe still would have to give it a couple of weeks and see how it's\nperforming in the long run because sometimes on the release day\nthe models perform quite good , but over time they kind of deteriorate\nin their quality .\nSo , we hope DeepSeek doesn't do that .\nHistorically , they haven't done that , but let's just see .\nJust going to put it out there because right now we're basing this\noff of benchmarks .\nBut that's it for today's video .\nMake sure you guys are subscribed to the channel .\nFollow our new newsletter as well at universeofai .\nbehiiv .\ncom as well as subscribe to the main channel World of AI and support\nus on X by following the Universe of AIZ as well .\nUntil then , I'll see you guys in the next", "language": "en", "duration": 626.126063, "segments": [ { "start": 0.0, "end": 3.499, "text": "So , Deepseek version 4 Pro is officially out today .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 3.509, "end": 6.21, "text": "Now , you might be confused because this model was technically", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 6.22, "end": 9.56, "text": "out , but it was in general availability , but this is the official", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 9.57, "end": 12.431, "text": "release , meaning that they have trained the model a little bit", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 12.441, "end": 15.861, "text": "more and produced a stronger version of the model that is available", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 15.871, "end": 16.34, "text": "today .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 16.35, "end": 19.087, "text": "Now , if I were to summarize today's release in one simple sentence", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 19.097, "end": 22.401, "text": ", it is that the price to performance is becoming a very ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 22.411, "end": 25.432, "text": "very important thing because not only do we have a new model from", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 25.442, "end": 28.824, "text": "the Deepseek team , the SpaceX team also dropped Grock 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 28.834, "end": 29.39, "text": "6 six .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 29.4, "end": 32.813, "text": "And both of these models are competing with the best Frontier Labs", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 32.823, "end": 34.816, "text": "at a fraction of their cost .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 34.826, "end": 36.181, "text": "Let's start with Gro 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 36.191, "end": 36.577, "text": "6 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 36.587, "end": 39.511, "text": "And one thing I'm going to say is that I'm genuinely surprised", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 39.521, "end": 40.779, "text": "by the SpaceX team .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 40.789, "end": 43.158, "text": "I'm not trying to glaze them or Elon Musk or anything .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 43.168, "end": 45.061, "text": "I was just very critical of this lab .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 45.071, "end": 48.421, "text": "In 2025 , they dropped Grog 4 and other models like that ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 48.431, "end": 52.121, "text": "but I wasn't really , you know , mind blown with their performance", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 52.131, "end": 55.15, "text": "because they're pretty they're pretty subpar compared to any of", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 55.16, "end": 56.493, "text": "the other labs out there .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 56.503, "end": 59.89, "text": "But in 2026 , it looks like things have kind of changed a little", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 59.9, "end": 60.962, "text": "because Grock 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 60.972, "end": 64.732, "text": "5 was quite competitive based off what it cost and Grock 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 64.742, "end": 67.018, "text": "6 is actually not too bad .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 67.028, "end": 69.49, "text": "If we take a look at the benchmarks , for example ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 69.5, "end": 72.998, "text": "if we start with the artificial analysis intelligence index ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 73.008, "end": 77.383, "text": "this model achieves a 61 and Fable 5 is at 62 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 77.393, "end": 80.333, "text": "Now , what's really important to remember is that once again ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 80.343, "end": 83.124, "text": "this model is quite cheap compared to Fable 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 83.134, "end": 87.35, "text": "This is about 2 per million input tokens and 6 per million output", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 87.36, "end": 87.995, "text": "tokens .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 88.005, "end": 92.594, "text": "In Fable 5 sits at 10 per million input tokens and 50 per million", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 92.604, "end": 93.643, "text": "output tokens .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 93.653, "end": 96.709, "text": "And this is why you start to appreciate this release a little bit", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 96.719, "end": 97.113, "text": "more .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 97.123, "end": 100.29, "text": "It might not beat the performance of the best models is matching", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 100.3, "end": 100.787, "text": "them .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 100.797, "end": 102.14, "text": "Even GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "Even GBT 5 .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 102.15, "end": 105.911, "text": "6 so which is a pretty capable model and this is set at max on", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 105.921, "end": 107.876, "text": "the artificial analysis index .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 107.886, "end": 110.578, "text": "This model achieves 61 and Grok 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "This model achieves 61 and Grock 4 .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 110.588, "end": 111.382, "text": "6 61 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 111.392, "end": 113.554, "text": "So it ties it and it's much cheaper .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 113.564, "end": 115.966, "text": "And then even on all of these other benchmarks ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 115.976, "end": 119.093, "text": "for example , the Code Val one , it actually beats Fable 5 ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "for example , the GDP Val one , it actually beats Fable 5 ,", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 119.103, "end": 120.936, "text": "which is at 1741 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 120.946, "end": 122.177, "text": "And then GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 122.187, "end": 124.059, "text": "6 , it's at 1728 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 124.069, "end": 126.463, "text": "I'm not sure why they didn't choose Opus 5 as well ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 126.473, "end": 129.666, "text": "but I guess they wanted to choose the quote unquote strongest model", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 129.676, "end": 130.709, "text": "lineup from each lab .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 130.719, "end": 133.653, "text": "And they chose Fable 5 for Enthropic , which is fair .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 133.663, "end": 136.667, "text": "And then Deep Software Engineering one , which is a critical benchmark", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 136.677, "end": 137.143, "text": ".", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 137.153, "end": 140.335, "text": "This model doesn't beat Fable 5 or GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 140.345, "end": 142.072, "text": "6 six soul , but it gets close to it .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 142.082, "end": 143.005, "text": "It's 65 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 143.015, "end": 145.589, "text": "9 and Fable 5 sits at 70 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 145.599, "end": 148.306, "text": "But if you're getting results that are pretty close and the model", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 148.316, "end": 151.182, "text": "is five times cheaper , I wouldn't be too disappointed with this", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 151.192, "end": 151.741, "text": "result .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 151.751, "end": 153.818, "text": "And then same thing with Cursor Bench 3 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 153.828, "end": 155.971, "text": "2 , the model achieves 69 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 155.981, "end": 157.654, "text": "9 , funny number .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 157.664, "end": 159.367, "text": "And Fable 5 sits at 70 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 159.377, "end": 159.891, "text": "5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 159.901, "end": 161.01, "text": "So once again , closer .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 161.02, "end": 161.949, "text": "And Grock 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 161.959, "end": 163.443, "text": "6 beats GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 163.453, "end": 164.61, "text": "6 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "6 Soul .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 164.62, "end": 166.125, "text": "Same thing with the Frontier Code .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 166.135, "end": 169.037, "text": "It gets close to Fable 5 , beats GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 169.047, "end": 169.737, "text": "6 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "6 Soul .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 169.747, "end": 172.558, "text": "So yeah , this model is actually available in cursor .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 172.568, "end": 176.265, "text": "So the partnership with cursor or I guess the acquisition has really", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 176.275, "end": 180.29, "text": "helped SpaceX make some strides in the AI space this year and Grock", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 180.3, "end": 183.49, "text": "build is something that you know maybe not a lot of us have been", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 183.5, "end": 187.091, "text": "using so far but it's probably going to be another platform like", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 187.101, "end": 190.543, "text": "codeex or cloud code that we start to use but obviously cursor", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 190.553, "end": 191.747, "text": "is quite strong as well .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 191.757, "end": 195.119, "text": "So you have options available for you to use them in both .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 195.129, "end": 198.41, "text": "And one thing to note is that they're offering two times usage", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 198.42, "end": 201.037, "text": "inside Grok Build and Cursor for the first week .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "inside Grock build and cursor for the first week .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 201.047, "end": 203.742, "text": "So if you just want to try it out , see what you feel about it", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 203.752, "end": 206.939, "text": ", then you know it might be worth trying it out right now cuz you", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 206.949, "end": 209.054, "text": "get double the usage in the first week .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 209.064, "end": 211.657, "text": "Now if we were to take a look at some of the outputs that people", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 211.667, "end": 213.666, "text": "have been generating with Grok 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "have been generating with Grock 4 .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 213.676, "end": 217.574, "text": "6 , what we're looking at right now is a Falcon 9 booster return", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 217.584, "end": 218.689, "text": "sequence simulation .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 218.699, "end": 221.8, "text": "And this was done in a single HTML file .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 221.81, "end": 225.229, "text": "And as I said , if you expected to get this type of output from", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 225.239, "end": 229.602, "text": "Grock in 2025 , you would be kind of surprised because you wouldn't", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 229.612, "end": 231.99, "text": "expect something like this to be generated with Grock .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 232.0, "end": 234.708, "text": "But now it looks like we have to start taking the Grok team a", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "But now it looks like we have to start taking the Grock team a", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 234.718, "end": 237.945, "text": "little bit more serious because this output is quite competitive", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 237.955, "end": 238.384, "text": ".", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 238.394, "end": 241.1, "text": "And obviously , we're just looking at a simulation and we're just", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 241.11, "end": 243.737, "text": "basing it off of a visual representation .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 243.747, "end": 246.374, "text": "But if the model is able to produce something like this consistently", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 246.384, "end": 250.275, "text": ", then I would expect a lot of people to start adopting Grock because", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 250.285, "end": 253.118, "text": "number one , it is cheaper than the other labs at least at the", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 253.128, "end": 255.645, "text": "moment because we don't know if this pricing strategy is going", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 255.655, "end": 258.645, "text": "to be sustainable for the SpaceX team in the long run .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 258.655, "end": 261.409, "text": "But at least for now , their models are definitely cheaper compared", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 261.419, "end": 262.358, "text": "to the others .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 262.368, "end": 265.753, "text": "And as I mentioned , this model excelled at the artificial analysis", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 265.763, "end": 266.385, "text": "index .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 266.395, "end": 269.781, "text": "This model jumped to probably number four model .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 269.791, "end": 271.93, "text": "It's tied to GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 271.94, "end": 273.12, "text": "6 pretty much similar .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 273.13, "end": 276.329, "text": "So you could say number three as well , but the models before that", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 276.339, "end": 280.646, "text": "are Fable 5 and Opus 5 , which are only above the model by about", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 280.656, "end": 283.204, "text": "a 1 or a 2 difference .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 283.214, "end": 286.242, "text": "So yeah , even on this intelligence index , which if you're not", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 286.252, "end": 288.32, "text": "familiar with has nine evaluations .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 288.33, "end": 292.396, "text": "So on all of these evaluations , it's kind of matching almost Fable", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 292.406, "end": 294.796, "text": "5 performance , which is crazy to see .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 294.806, "end": 298.153, "text": "Before we continue , we just launched the Universe of AI newsletter", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 298.163, "end": 298.552, "text": ".", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 298.562, "end": 301.511, "text": "If you want to stay on top of AI news without having to hunt for", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 301.521, "end": 303.345, "text": "it , link is in the description .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 303.355, "end": 304.46, "text": "Don't miss out .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 304.47, "end": 307.55, "text": "And what you see on screen right now is a racing game that Grok", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "And what you see on screen right now is a racing game that Grock", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 307.56, "end": 308.07, "text": "4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 308.08, "end": 308.877, "text": "6 build .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 308.887, "end": 311.918, "text": "And based off of this post , the model took about 1 minute and", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 311.928, "end": 314.941, "text": "it was a fiveword prompt , which was a create a simple racing game", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 314.951, "end": 316.019, "text": "in HTML .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 316.029, "end": 319.117, "text": "So , if you're able to generate something like this easily using", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 319.127, "end": 319.85, "text": "Grock 4 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 319.86, "end": 322.05, "text": "6 , I think a lot of people will be happy .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 322.06, "end": 326.205, "text": "And this is a more detailed analysis of what it costs to run GPT", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 326.215, "end": 326.535, "text": "5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 326.545, "end": 330.022, "text": "6 6 on the artificial analysis index and what it produced meaning", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 330.032, "end": 330.974, "text": "the output .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 330.984, "end": 334.463, "text": "Both of these models if you remember scored 61 on the artificial", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 334.473, "end": 335.495, "text": "analysis index .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 335.505, "end": 337.98, "text": "Now to run the whole test with GPT 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 337.99, "end": 339.885, "text": "6 it cost about 2 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 339.895, "end": 341.758, "text": "6 it cost about 2 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "8K and then with Grock 4 .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 341.768, "end": 343.37, "text": "6 it cost about 1 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 343.38, "end": 344.05, "text": "1Kish .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 344.06, "end": 347.215, "text": "And this tells you that you're getting similar level of performance", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 347.225, "end": 348.269, "text": "at half the cost .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 348.279, "end": 351.643, "text": "So yeah , this is a big release for the SpaceX team because they", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 351.653, "end": 354.529, "text": "just proved once again that they are a lab that you seriously start", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 354.539, "end": 356.83, "text": "need to considering , especially in 2026 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 356.84, "end": 360.574, "text": "And I'm going to talk more about Deep Seek version 4 Pro GA .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 360.584, "end": 364.1, "text": "But basically what we're seeing today is that both of these releases", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 364.11, "end": 367.869, "text": "kind of emphasize the fact that performance and all above that", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 367.879, "end": 370.755, "text": "is the price at what you're getting for that performance is becoming", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 370.765, "end": 374.362, "text": "more and more important for all users because we see many labs", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 374.372, "end": 378.129, "text": "now focusing on creating the best model at the cheapest cost .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 378.139, "end": 381.657, "text": "Last year in 2025 most of the labs were just focused on I would", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 381.667, "end": 384.222, "text": "say creating the strongest model .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 384.232, "end": 387.269, "text": "Yes , cost was important , but I think most of the times the frontier", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 387.279, "end": 390.579, "text": "labs , meaning OpenAI , Anthropic , were kind of more lenient on", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "labs , meaning OpenAI , Enthropic , were kind of more lenient on", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 390.589, "end": 393.85, "text": "that fact because they didn't have as strong of a competition .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 393.86, "end": 398.483, "text": "Intelligent models that are maybe not always ahead of OpenAI Enthropic", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 398.493, "end": 401.197, "text": ", but match their performance at a fraction of the cost .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 401.207, "end": 405.119, "text": "So yes , price to performance ratio is becoming a critical I would", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "So yes , price toerformance ratio is becoming a critical I would", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 405.129, "end": 407.705, "text": "say indicator in 2026 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 407.715, "end": 410.611, "text": "Now , this is the updated benchmark chart after the release of", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 410.621, "end": 411.492, "text": "the new model .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 411.502, "end": 414.373, "text": "And one thing you'll see across the board is that it matches the", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 414.383, "end": 416.694, "text": "top level performance of many of the models .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 416.704, "end": 419.095, "text": "The one thing interesting over here is that they haven't put Opus", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 419.105, "end": 420.454, "text": "5 here for some reason .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 420.464, "end": 423.497, "text": "There is Fable 5 here that we can compare this model against .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 423.507, "end": 425.818, "text": "But one thing you'll notice is that DeepSeek ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "But one thing you'll notice is that Deep Seek ,", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 425.828, "end": 428.085, "text": "remember this model costs 43 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 428.095, "end": 432.541, "text": "5 cents per million input tokens and 87 cents per million output", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 432.551, "end": 433.261, "text": "tokens .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 433.271, "end": 436.141, "text": "While the other models all over here are way more expensive than", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 436.151, "end": 436.622, "text": "that .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 436.632, "end": 439.29, "text": "So the first thing if you look at terminal bench 2 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 439.3, "end": 441.379, "text": "1 the model scores 87 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 441.389, "end": 442.144, "text": "9 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 442.154, "end": 444.406, "text": "The older version of the model was 72 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 444.416, "end": 448.307, "text": "1 and the flash version which we got last week was 82 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 448.317, "end": 449.027, "text": "7 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 449.037, "end": 451.989, "text": "And what's crazy is that Fable 5 is 88 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 451.999, "end": 453.029, "text": "Yes , 88 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 453.039, "end": 454.637, "text": "So this model is only .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 454.647, "end": 456.577, "text": "1 behind Fable 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 456.587, "end": 459.05, "text": "And then on the Cyber Gym , which is Cyber Security ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 459.06, "end": 461.379, "text": "the model actually beats Fable 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 461.389, "end": 463.309, "text": "Fable 5 sits at 83 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 463.319, "end": 466.433, "text": "1 while Deepseek version 4 Pro , the new one that we got today", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 466.443, "end": 467.794, "text": "3 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": ", is at 83 .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 467.804, "end": 468.515, "text": "3 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 468.525, "end": 471.716, "text": "Now , if this tells you something is that this model is once again", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 471.726, "end": 473.557, "text": "really geared at cyber security .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 473.567, "end": 476.6, "text": "We saw the Flash model also be geared towards cyber security and", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 476.61, "end": 479.162, "text": "becoming a model that was quite capable in that area .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 479.172, "end": 482.284, "text": "And we're seeing the same thing today with the new version 4 Pro", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 482.294, "end": 483.725, "text": "which sits at 83 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 483.735, "end": 486.608, "text": "3 outperforming Fable 5 on the benchmark .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 486.618, "end": 490.129, "text": "On deep software engineering , this model is not beating Fable", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 490.139, "end": 494.247, "text": "5 because Fable 5 sits at 70 , but the model achieves 62 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 494.257, "end": 496.494, "text": "7 which is ahead of GLM 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 496.504, "end": 499.959, "text": "2 and a bit behind Kimmy K3 which sits at 67 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 499.969, "end": 500.727, "text": "5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 500.737, "end": 503.865, "text": "But one thing again , this is a big jump compared to the preview", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 503.875, "end": 506.036, "text": "version of the model which was at 12 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 506.046, "end": 506.519, "text": "8 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 506.529, "end": 509.405, "text": "So yeah , they have really trained this model and they have improved", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 509.415, "end": 512.748, "text": "it on the back end because we can clearly see in this deep software", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 512.758, "end": 514.18, "text": "engineering benchmark .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 514.19, "end": 516.966, "text": "And there's a couple of other benchmarks , but one key benchmark", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 516.976, "end": 521.103, "text": "that this model excels at is the automation bench where the model", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 521.113, "end": 522.43, "text": "achieves 31 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 522.44, "end": 524.525, "text": "8 and Fable 5 29 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 524.535, "end": 525.082, "text": "1 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 525.092, "end": 527.789, "text": "So once again , the model outperforms Fable 5 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 527.799, "end": 530.018, "text": "Now this is a fraction of the cost as I mentioned .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 530.028, "end": 534.128, "text": "This is 43 cents and 87 for input and output respectively versus", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 534.138, "end": 536.05, "text": "Fable 5 which is at 10 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 536.06, "end": 536.833, "text": "50 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 536.843, "end": 539.537, "text": "So yes , if we look at the terminal bench for example ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 539.547, "end": 542.405, "text": "we are seeing this model achieve a result which is 0 .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 542.415, "end": 546.828, "text": "1 behind the best model out there Fable 5 and the pricing is insane", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 546.838, "end": 553.934, "text": "to look at 43 versus 10 87 versus 50 per million input tokens .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 553.944, "end": 558.424, "text": "So when we turn that into a fraction , this model is 57 times cheaper", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 558.434, "end": 558.824, "text": ".", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 558.834, "end": 562.035, "text": "And earlier last week , I think I made a video about how DeepSeek", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 562.045, "end": 565.49, "text": "version 4 Pro was going to achieve a result like this because there", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 565.5, "end": 567.245, "text": "was a investor report that leaked .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 567.255, "end": 569.877, "text": "And when I saw the pricing where it said that it was going to match", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 569.887, "end": 574.22, "text": "Fable 5 or even outperform it and be at 57 times cheaper per cost", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 574.23, "end": 574.544, "text": ".", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 574.554, "end": 577.027, "text": "When I was reading the investor report , I didn't really believe", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 577.037, "end": 577.587, "text": "it .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 577.597, "end": 580.383, "text": "But today , it is proven because at least on the benchmark so far", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 580.393, "end": 583.579, "text": ", yes , I'm just saying the benchmarks , we are seeing similar", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 583.589, "end": 584.698, "text": "level performance .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 584.708, "end": 588.053, "text": "Now , if this model actually performs like that in production ,", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 588.063, "end": 589.891, "text": "it's a little too early to tell yet .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 589.901, "end": 592.513, "text": "We still would have to give it a couple of weeks and see how it's", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 592.523, "end": 595.617, "text": "performing in the long run because sometimes on the release day", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 595.627, "end": 599.09, "text": "the models perform quite good , but over time they kind of deteriorate", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 599.1, "end": 600.13, "text": "in their quality .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 600.14, "end": 601.755, "text": "So , we hope DeepSeek doesn't do that .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "So , we hope DeepS doesn't do that .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 601.765, "end": 604.219, "text": "Historically , they haven't done that , but let's just see .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 604.229, "end": 606.612, "text": "Just going to put it out there because right now we're basing this", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 606.622, "end": 608.279, "text": "off of benchmarks .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 608.289, "end": 609.801, "text": "But that's it for today's video .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 609.811, "end": 611.831, "text": "Make sure you guys are subscribed to the channel .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 611.841, "end": 615.119, "text": "Follow our new newsletter as well at universeofai .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 615.129, "end": 615.506, "text": "behiiv .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow", "text_original": "behive .", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "start": 615.516, "end": 618.837, "text": "com as well as subscribe to the main channel World of AI and support", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 618.847, "end": 621.813, "text": "us on X by following the Universe of AIZ as well .", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" }, { "start": 621.823, "end": 625.409, "text": "Until then , I'll see you guys in the next", "source": "source_caption", "timing_source": "vad_whisper_monotonic_reflow" } ], "transcript_selection": { "schema": "videonote.transcript_selection.v1", "status": "SELECTED", "requested_language": "en", "duration": 626.126063, "reference_path": "/Users/sagawa/AI_WORK/video_notes/url/20260813_143826__DeepSeek_V4_Pro_Is_Out_Today_And_It_s_A_Problem_For_Every_AI_Lab/reference_subtitle.vtt", "reference_kind": "automatic_original", "reference_language": "en-orig", "selected_source": "youtube_auto_retimed_with_whisper_anchors", "decision_reason": "automatic_caption_cumulative_drift_repaired_by_monotonic_anchors", "whisper_rejected": false, "source_profile": { "segment_count": 292, "multilingual": false, "first_start": 0.08, "final_end": 627.519, "duration_coverage": 1.0, "maximum_gap_seconds": 0.08, "inverted_count": 0, "language": { "requested": "en", "matched": true, "latin": 8426, "han": 0, "kana": 0, "hangul": 0 }, "hallucination_region_count": 0, "hallucination_regions": [], "usable_full": true, "usable_partial": true }, "whisper_profile": { "segment_count": 115, "multilingual": false, "first_start": 0.0, "final_end": 625.706, "duration_coverage": 0.999329, "maximum_gap_seconds": 0.5, "inverted_count": 0, "language": { "requested": "en", "matched": true, "latin": 8473, "han": 0, "kana": 0, "hangul": 0 }, "hallucination_region_count": 0, "hallucination_regions": [], "usable_full": true, "usable_partial": true }, "comparison": { "window_seconds": 30.0, "compared_window_count": 21, "median_score": 0.9355, "divergent_window_count": 0, "divergent_ratio": 0.0, "divergent_windows": [] }, "timestamp_alignment": { "applied": false, "estimated_offset_seconds": 0.0, "matched_segment_count": 0, "match_scores": [] }, "caption_timing_repair": { "method": "monotonic_multi_anchor_piecewise_interpolation", "applied": true, "anchor_count": 42, "anchors": [ { "source_time": 6.16, "whisper_time": 7.6, "score": 0.884 }, { "source_time": 18.84, "whisper_time": 22.93, "score": 0.7503 }, { "source_time": 32.12, "whisper_time": 38.26, "score": 0.5769 }, { "source_time": 58.919, "whisper_time": 53.96, "score": 0.6997 }, { "source_time": 71.88, "whisper_time": 70.88, "score": 0.7914 }, { "source_time": 85.2, "whisper_time": 88.39, "score": 0.7154 }, { "source_time": 98.6, "whisper_time": 102.12, "score": 0.6634 }, { "source_time": 117.84, "whisper_time": 113.98, "score": 0.8307 }, { "source_time": 128.959, "whisper_time": 128.28, "score": 0.812 }, { "source_time": 139.92, "whisper_time": 142.59, "score": 0.6631 }, { "source_time": 153.08, "whisper_time": 155.64, "score": 0.7572 }, { "source_time": 166.68, "whisper_time": 170.6, "score": 0.6845 }, { "source_time": 179.44, "whisper_time": 184.22, "score": 0.5983 }, { "source_time": 192.079, "whisper_time": 197.4, "score": 0.6432 }, { "source_time": 218.44, "whisper_time": 212.75, "score": 0.6284 }, { "source_time": 231.36, "whisper_time": 228.07, "score": 0.8275 }, { "source_time": 243.76, "whisper_time": 243.65, "score": 0.9251 }, { "source_time": 257.159, "whisper_time": 258.31, "score": 0.9215 }, { "source_time": 271.399, "whisper_time": 274.41, "score": 0.8034 }, { "source_time": 285.039, "whisper_time": 288.9, "score": 0.6959 }, { "source_time": 297.56, "whisper_time": 302.6, "score": 0.6689 }, { "source_time": 323.279, "whisper_time": 324.2, "score": 0.8869 }, { "source_time": 336.159, "whisper_time": 338.81, "score": 0.7413 }, { "source_time": 348.4, "whisper_time": 352.7, "score": 0.7468 }, { "source_time": 373.48, "whisper_time": 367.89, "score": 0.6816 }, { "source_time": 389.2, "whisper_time": 383.75, "score": 0.6494 }, { "source_time": 401.32, "whisper_time": 400.71, "score": 0.8621 }, { "source_time": 414.639, "whisper_time": 416.38, "score": 0.8647 }, { "source_time": 427.279, "whisper_time": 431.55, "score": 0.6719 }, { "source_time": 449.84, "whisper_time": 448.643, "score": 0.8843 }, { "source_time": 463.52, "whisper_time": 463.346, "score": 0.9822 }, { "source_time": 479.799, "whisper_time": 477.266, "score": 0.8392 }, { "source_time": 492.24, "whisper_time": 491.386, "score": 0.9554 }, { "source_time": 505.76, "whisper_time": 506.246, "score": 0.9218 }, { "source_time": 519.32, "whisper_time": 521.526, "score": 0.8263 }, { "source_time": 543.519, "whisper_time": 536.726, "score": 0.6771 }, { "source_time": 555.84, "whisper_time": 552.276, "score": 0.6806 }, { "source_time": 568.8, "whisper_time": 566.826, "score": 0.8673 }, { "source_time": 582.199, "whisper_time": 581.726, "score": 0.9453 }, { "source_time": 594.6, "whisper_time": 595.746, "score": 0.8744 }, { "source_time": 606.88, "whisper_time": 610.076, "score": 0.8034 }, { "source_time": 618.639, "whisper_time": 621.716, "score": 0.6339 } ], "source_block_count": 52, "whisper_block_count": 43, "matched_block_count": 42, "case_normalization": { "applied": false, "changed_segment_count": 0 }, "source_text_preserved": true, "reason": "source_text_retimed_from_monotonic_whisper_anchors", "minimum_anchor_delta_seconds": -6.793, "maximum_anchor_delta_seconds": 6.14, "drift_span_seconds": 12.933 }, "filled_source_profile": { "segment_count": 53, "multilingual": false, "first_start": 0.249, "final_end": 626.126, "duration_coverage": 1.0, "maximum_gap_seconds": 0.249, "inverted_count": 1, "language": { "requested": "en", "matched": true, "latin": 8426, "han": 0, "kana": 0, "hangul": 0 }, "hallucination_region_count": 0, "hallucination_regions": [], "usable_full": false, "usable_partial": false }, "whisper_duration_guard": { "media_duration": 626.126, "rejected_segment_count": 0, "rejected_segments": [] } }, "caption_alignment": { "method": "vad_whisper_monotonic_reflow", "applied": true, "source_text_authority": "preserved" }, "presegmentation_correction": { "applied": true, "correction_count": 16 } }