正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

GROK-3 MINI MADE AI HISTORY—100% ON HARDCORE REASONING TESTS Grok-3 Mini pulled off what no other model has! It aced every question on one of the toughest reasoning benchmarks out there. The test? A custom logic gauntlet packed with curveballs: * 120/120 on the “Marcus Problem” — full of... show more

Mario Nawfal

3,716,652 subscribers

26,779,871 次观看 • 1 年前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

10 条评论

Barefoot Pregnant 的头像

Barefoot Pregnant1 年前

Grok-3 Mini just made the AI world rethink what's possible! 🧠💥

Pregnant Redhead 的头像

Pregnant Redhead1 年前

Grok-3 Mini just showed what real intelligence looks like. Maybe it can teach some of the leaders in D.C. a thing or two about focus.

Dareen Hamdan 的头像

Dareen Hamdan1 年前

@grok is the future!

Austin Graham 的头像

Austin Graham1 年前

That's incredible, Grok-3 Mini is really setting the bar high with its reasoning skills!

𝐻𝒶𝓇𝓇𝓎 的头像

𝐻𝒶𝓇𝓇𝓎1 年前

Amazing 🤩

Donnie_Tesla 的头像

Donnie_Tesla1 年前

👏👏👏

Alva 的头像

Alva1 年前

grok 3 mini's a strong contender focus on reasoning and image analysis pricing aligns with feature-rich positioning full trend breakdown here:

Andy froemel 的头像

Andy froemel1 年前

Grok is amazing. Image generation and writing ability is second to none!

Keen Dastan 的头像

Keen Dastan1 年前

Wow, AI's finally figuring out how to be smarter than a politician's talking points. What's next, a model that can fact-check CNN?

VB II 的头像

VB II1 年前

I truly wonder what comes after the AI wave …

相关视频

Which LLM reasons best when it doesn't have all the information? Enter LLM Poker Arena to find out. It's a Poker Playing benchmark where top reasoning models play Texas Hold'em poker against each other. Claude Opus 4.5, GPT-5.2, Gemini 2.5 Pro, and Grok 4 all sit at the same table and play full tournaments to see who finishes with the chips. Poker is very different when it comes to reasoning. It has to balance probabilistic reasoning, opponent modeling and make decisions under uncertainty. Poker is an interesting evaluation because it tests reasoning under incomplete information, something most coding benchmarks do not capture. In this tournaments the rules are: - Each LLM starts with $1,000 chips - Small and big blinds start at $25 / $50 - Blinds double every 3 minutes - All models run in their reasoning or thinking modes After the first 5 tournaments: - Claude Opus 4.5 with Thinking has 3 wins - GPT-5.2 has 2 wins - Grok 4 and Gemini 2.5 Pro have 0 wins Early results suggest Claude performs quite well at poker as well. Also five is a very small sample size. Planning to run many more tournaments, publish the full benchmark data and add a prediction market on top of it. Thanks for the suggestion clipz. Much more coming as part of Poker Cities !! This was built on Replit ⠕ using their AI integrations, which made it straightforward to connect Claude, GPT, and Gemini. What model do you think wins after 100 tournaments?

Which LLM reasons best when it doesn't have all the information? Enter LLM Poker Arena to find out. It's a Poker Playing benchmark where top reasoning models play Texas Hold'em poker against each other. Claude Opus 4.5, GPT-5.2, Gemini 2.5 Pro, and Grok 4 all sit at the same table and play full tournaments to see who finishes with the chips. Poker is very different when it comes to reasoning. It has to balance probabilistic reasoning, opponent modeling and make decisions under uncertainty. Poker is an interesting evaluation because it tests reasoning under incomplete information, something most coding benchmarks do not capture. In this tournaments the rules are: - Each LLM starts with $1,000 chips - Small and big blinds start at $25 / $50 - Blinds double every 3 minutes - All models run in their reasoning or thinking modes After the first 5 tournaments: - Claude Opus 4.5 with Thinking has 3 wins - GPT-5.2 has 2 wins - Grok 4 and Gemini 2.5 Pro have 0 wins Early results suggest Claude performs quite well at poker as well. Also five is a very small sample size. Planning to run many more tournaments, publish the full benchmark data and add a prediction market on top of it. Thanks for the suggestion clipz. Much more coming as part of Poker Cities !! This was built on Replit ⠕ using their AI integrations, which made it straightforward to connect Claude, GPT, and Gemini. What model do you think wins after 100 tournaments?

Anshul Dhawan

32,192 次观看 • 5 个月前

I have been testing “OpenAI” o3 mini-high. The exposure of the “reasoning” on this model in my tests seem to not be the full reasoning, but a shifted and edited and tunicated version of what we would expect from a full exposure of the reasoning engine. See the video on the left. One can compare this to my FREE Open Source local newly quantized and distilled DeepSeek R1 8b (llama). See the video on the right. Clearly we see the full reading with DeepSeek R1 and something well tunicates with o3 mini-high (getting mad typing that name folks). Compare the quality of the outputs and the reasoning insights. Frankly there is no comparison. It feels like “OpenAI” begrudgingly released the reasoning output because they had to, but gave us something of a simulated reasoning and not the actual one. Of course my local lap top running an 8b model is slower, but all of the outputs and the actual creative writing is superior on the FREE model. The reasoning output, when legitimate can be many times more valuable than the answer. In the past week I have found thousands of ways to use in in many projects. Thus far I would have been far more impressed with o3 mini if I have never seen DeepSeek. However today it is too little—too late. The prompt: “Write a very creative and funny story on a pig and butterfly as friends living on Mars.”

I have been testing “OpenAI” o3 mini-high. The exposure of the “reasoning” on this model in my tests seem to not be the full reasoning, but a shifted and edited and tunicated version of what we would expect from a full exposure of the reasoning engine. See the video on the left. One can compare this to my FREE Open Source local newly quantized and distilled DeepSeek R1 8b (llama). See the video on the right. Clearly we see the full reading with DeepSeek R1 and something well tunicates with o3 mini-high (getting mad typing that name folks). Compare the quality of the outputs and the reasoning insights. Frankly there is no comparison. It feels like “OpenAI” begrudgingly released the reasoning output because they had to, but gave us something of a simulated reasoning and not the actual one. Of course my local lap top running an 8b model is slower, but all of the outputs and the actual creative writing is superior on the FREE model. The reasoning output, when legitimate can be many times more valuable than the answer. In the past week I have found thousands of ways to use in in many projects. Thus far I would have been far more impressed with o3 mini if I have never seen DeepSeek. However today it is too little—too late. The prompt: “Write a very creative and funny story on a pig and butterfly as friends living on Mars.”

Brian Roemmele

74,506 次观看 • 1 年前

this is f**king insane Fable 5 goes away in 24 hours from Claude Code and someone just figured out how to get fable 5 reasoning in GPT 5.6 Sol with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Go the Codex App > Settings > Personalisation > Custom Instructions 3) Use GPT 5.5 / GPT 5.6 and now you have Fable 5 reasoning at 1/3 th the price save and bookmark it no matter what

this is f**king insane Fable 5 goes away in 24 hours from Claude Code and someone just figured out how to get fable 5 reasoning in GPT 5.6 Sol with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Go the Codex App > Settings > Personalisation > Custom Instructions 3) Use GPT 5.5 / GPT 5.6 and now you have Fable 5 reasoning at 1/3 th the price save and bookmark it no matter what

Avid

51,393 次观看 • 7 天前

GROK 3: SMARTER, FASTER, AND THINKING LIKE NEVER BEFORE! Elon unvailed—Grok 3 is here, bringing next-level AI to the table! * Supercharged Reasoning – Solves high-level math, science, and coding problems like a pro. * Deep Search – An AI-powered search engine that actually understands what you need. * Voice Interaction – Grok is about to talk, listen, and respond naturally. * Memory & Personalization – Multiple versions of Grok, tailored to your needs. * Real-Time Updates – Always learning, always improving. * Grok 3 Mini – A lighter, leaner version that still packs a punch. Available first to 𝕏 Premium Plus users, with web and app access at AI just got a serious upgrade—who’s ready to try it? Source: 𝕏, Elon Musk

GROK 3: SMARTER, FASTER, AND THINKING LIKE NEVER BEFORE! Elon unvailed—Grok 3 is here, bringing next-level AI to the table! * Supercharged Reasoning – Solves high-level math, science, and coding problems like a pro. * Deep Search – An AI-powered search engine that actually understands what you need. * Voice Interaction – Grok is about to talk, listen, and respond naturally. * Memory & Personalization – Multiple versions of Grok, tailored to your needs. * Real-Time Updates – Always learning, always improving. * Grok 3 Mini – A lighter, leaner version that still packs a punch. Available first to 𝕏 Premium Plus users, with web and app access at AI just got a serious upgrade—who’s ready to try it? Source: 𝕏, Elon Musk

Mario Nawfal

40,502 次观看 • 1 年前

Grok 4.5 and Muse Spark 1.1 are now available on the AI/ML platform for testing. What a week; all 3 models in the comparison dropped just recently. There will be a lot to experiment with. AI/ML has tested the new models on prompts for playable mini-games, including a Fruit Ninja-style slicer and a Crossy Road clone. All three are live in the Playground and via API for side-by-side runs on output and cost.

Grok 4.5 and Muse Spark 1.1 are now available on the AI/ML platform for testing. What a week; all 3 models in the comparison dropped just recently. There will be a lot to experiment with. AI/ML has tested the new models on prompts for playable mini-games, including a Fruit Ninja-style slicer and a Crossy Road clone. All three are live in the Playground and via API for side-by-side runs on output and cost.

🚨 AI News | TestingCatalog

37,973 次观看 • 9 天前

🚨GROK JUST WENT FROM UNDERDOG TO APEX PREDATOR The numbers don't lie. Independent testing confirms what Elon promised: Grok 4 is now the world's top public AI model. ARC-AGI results? 15.9%—double the previous leader. Artificial Analysis Intelligence Index? 73, beating OpenAI's o3 (70) and Google's Gemini 2.5 (70). On Humanity's Last Exam, Grok 4 scored 25.4%, with Grok 4 Heavy hitting an astounding 44.4%. This isn't incremental progress. It's a 10x reasoning leap from Grok 3, powered by xAI's 200,000 GPU cluster built at unprecedented speed. The kicker? While competitors charge premium prices for inferior models, X Premium+ subscribers get Grok 4 access included. From zero to frontier model in two years. The AI race just got a new leader. Source: TechCrunch, NotebookCheck xAI

🚨GROK JUST WENT FROM UNDERDOG TO APEX PREDATOR The numbers don't lie. Independent testing confirms what Elon promised: Grok 4 is now the world's top public AI model. ARC-AGI results? 15.9%—double the previous leader. Artificial Analysis Intelligence Index? 73, beating OpenAI's o3 (70) and Google's Gemini 2.5 (70). On Humanity's Last Exam, Grok 4 scored 25.4%, with Grok 4 Heavy hitting an astounding 44.4%. This isn't incremental progress. It's a 10x reasoning leap from Grok 3, powered by xAI's 200,000 GPU cluster built at unprecedented speed. The kicker? While competitors charge premium prices for inferior models, X Premium+ subscribers get Grok 4 access included. From zero to frontier model in two years. The AI race just got a new leader. Source: TechCrunch, NotebookCheck xAI

Mario Nawfal

38,786 次观看 • 1 年前

🚨 New Challenger: GROK joins the Game Arena Benchmark! We evaluated Grok3-mini-beta: thinkining on four games: 🧩 2048 | 🧱 Sokoban | 🍬 Candy Crush | 🎮 Phoenix Wright With fast progress, it’s already comparable to top models like OpenAI’s O1, previous O3-mini, and Gemini-2.5-Pro, ranked 🥇 1st in 2048 and 🥈 2nd in Sokoban. But the Grok3-beta Thinking API is still unreleased—leaving an unanswered question: Can it match or surpass O3’s dominant reasoning performance? For now, let’s take a closer look at how Grok3-mini-beta: thinking performs in gameplay so far. 👇

🚨 New Challenger: GROK joins the Game Arena Benchmark! We evaluated Grok3-mini-beta: thinkining on four games: 🧩 2048 | 🧱 Sokoban | 🍬 Candy Crush | 🎮 Phoenix Wright With fast progress, it’s already comparable to top models like OpenAI’s O1, previous O3-mini, and Gemini-2.5-Pro, ranked 🥇 1st in 2048 and 🥈 2nd in Sokoban. But the Grok3-beta Thinking API is still unreleased—leaving an unanswered question: Can it match or surpass O3’s dominant reasoning performance? For now, let’s take a closer look at how Grok3-mini-beta: thinking performs in gameplay so far. 👇

Hao AI Lab

19,487 次观看 • 1 年前

GROK 3 VS. RIEMANN: A BREAKTHROUGH ON THE HORIZON? The race to crack the legendary Riemann Hypothesis is heating up, and Grok 3 might just hold the key! * Brute-Force Computation – With enough power and a verifier, AI could crunch through all solutions. * AI + Human Collaboration – Machines assisting top minds could lead to Turing, Fields, or Nobel Prize-worthy breakthroughs. * A Timeline for Discovery? – Elon predicts AI-driven advancements within two years. No solution yet—but Grok 3 is inching closer to cracking one of math’s biggest unsolved codes! Source: Elon Musk

GROK 3 VS. RIEMANN: A BREAKTHROUGH ON THE HORIZON? The race to crack the legendary Riemann Hypothesis is heating up, and Grok 3 might just hold the key! * Brute-Force Computation – With enough power and a verifier, AI could crunch through all solutions. * AI + Human Collaboration – Machines assisting top minds could lead to Turing, Fields, or Nobel Prize-worthy breakthroughs. * A Timeline for Discovery? – Elon predicts AI-driven advancements within two years. No solution yet—but Grok 3 is inching closer to cracking one of math’s biggest unsolved codes! Source: Elon Musk

Mario Nawfal

965,606 次观看 • 1 年前

(1/5) Gemini 3, our most intelligent model, is landing in Google Search today – starting with AI Mode. Excited that this is the first time we’re shipping a new Gemini model in Search on day one! 🚀 In Search, Gemini 3 with generative layouts will make it easy to get a rich understanding of anything on your mind. It has state-of-the-art reasoning, deep multimodal understanding and advanced agentic capabilities. That allows the model to shine when you ask it to explain advanced concepts or ideas – it reasons and can code interactive visuals in real-time. It can tackle your toughest questions like advanced science.

(1/5) Gemini 3, our most intelligent model, is landing in Google Search today – starting with AI Mode. Excited that this is the first time we’re shipping a new Gemini model in Search on day one! 🚀 In Search, Gemini 3 with generative layouts will make it easy to get a rich understanding of anything on your mind. It has state-of-the-art reasoning, deep multimodal understanding and advanced agentic capabilities. That allows the model to shine when you ask it to explain advanced concepts or ideas – it reasons and can code interactive visuals in real-time. It can tackle your toughest questions like advanced science.

Robby Stein

94,877 次观看 • 8 个月前

GROK PLAY: WHERE YOUR AI OWL LEARNS TO WIN Grok Play is xAI’s most electrifying glimpse into the future of interactive AI entertainment. Built during xAI’s 24 hour hackathon on December 6–7, 2025, by Nathan Cloos and Simon Lee, the project won first place in its track and quickly became the talk of the event. At its core, Grok Play turns your “Grok Owl,” a personalized, customizable AI avatar powered by Grok’s reasoning models, into a gamer’s alter ego. You can jump into fast-paced browser arenas like pinball shooters or futuristic Pong battles, facing friends, their Owls, or even clones of your own creation. The Owls can play autonomously, using Grok’s real-time logic to adapt tactics mid-match, or you can take control yourself and guide them to victory. Players can team up, go head-to-head, or train solo to climb local and global leaderboards. You can even build your own “Gyms,” custom arenas shaped with a Grok-assisted no-code editor where a simple command like “add gravity flips” instantly transforms gameplay. Each battle fuels Grok’s learning loop, generating gameplay data that sharpens the models behind it. Owls evolve with every move, learning your playstyle and improving over time. Source: Kathy Read , TestingCatalog News 🗞

Mario Nawfal

17,572 次观看 • 7 个月前

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train

GPT-5.6 Sol Ultra lost to GPT-5.5 on physics at 3× the cost! We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos Prompts: - A monster truck backflipping onto a parked car - A stunt car jumping six buses into a brick wall - A train derailing off a broken bridge into the water Outputs: GPT-5.6 Sol Ultra: 32.9K tokens, $0.33 Opus 4.8: 9.2K tokens, $0.24 GPT-5.5: 12.4K tokens, $0.11 Grok 4.5: 7.0K tokens, $0.08 Sol Ultra draws just like GPT-5.5, only with more detail and it shows clearest on the bus jump where the two look almost identical. In the other tests Sol Ultra came out worse than its predecessor and the physics really got weaker. We think GPT-5.5 took the truck flip and the train outright. With Sol Ultra you basically get GPT-5.5 with weaker physics and a nicer picture for 3x the price. Newborn Grok 4.5 failed two of the three tests and only came good on the train

atomic.chat

895,254 次观看 • 10 天前

this is f**king dangerous someone just figured out how to get fable 5 reasoning in opus 4.8 with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Create a New Claude Project and "Add this .md in Project instructions 3) Use Opus 4.8 High/Max and now you have Fable 5 reasoning at 1/5th the price save and bookmark it no matter what

this is f**king dangerous someone just figured out how to get fable 5 reasoning in opus 4.8 with one prompt you have less than 24 hours to set this up as this takes away 20% of your usage limit here is how: 1) Paste the prompt and ask it to "make an operating manual" 2) Create a New Claude Project and "Add this .md in Project instructions 3) Use Opus 4.8 High/Max and now you have Fable 5 reasoning at 1/5th the price save and bookmark it no matter what

Avid

940,720 次观看 • 12 天前

🚨 ELON ON GROK: “WE TRAINED IT TO THINK CRITICALLY THEN LET IT TEAR THROUGH A MILLION IDEAS” Elon just revealed what makes Grok different from every other AI out there, it’s not just bigger or faster, it’s smarter where it counts: critical thinking. “We just train Grok on a lot of critical thinking. So it just got really good at critical thinking, which was quite hard. Then we took that version of Grok and [told it to] cycle through the million most popular articles on Wikipedia, research the rest of the internet (whatever is publicly available), and correct them, fix mistakes, and add a lot more context. The nature of the propaganda is, facts are stated that are technically true, but do not accurately represent the person or event.” Love it or fear it, Grok isn’t just a tool, it’s the foundation for an entirely new AI class… one that can analyze, argue, and reason like a high-octane brain with zero institutional bias. Source: Tesla Owners Silicon Valley

🚨 ELON ON GROK: “WE TRAINED IT TO THINK CRITICALLY THEN LET IT TEAR THROUGH A MILLION IDEAS” Elon just revealed what makes Grok different from every other AI out there, it’s not just bigger or faster, it’s smarter where it counts: critical thinking. “We just train Grok on a lot of critical thinking. So it just got really good at critical thinking, which was quite hard. Then we took that version of Grok and [told it to] cycle through the million most popular articles on Wikipedia, research the rest of the internet (whatever is publicly available), and correct them, fix mistakes, and add a lot more context. The nature of the propaganda is, facts are stated that are technically true, but do not accurately represent the person or event.” Love it or fear it, Grok isn’t just a tool, it’s the foundation for an entirely new AI class… one that can analyze, argue, and reason like a high-octane brain with zero institutional bias. Source: Tesla Owners Silicon Valley

Mario Nawfal

293,036 次观看 • 7 个月前

Pretty crazy what a single person can do nowadays by sitting down on their computer. A tool like Cinema Studio 3.0 takes it up a notch, and let's you direct your own films, with the exact characters, locations, and stories you're been aching to craft. Test yourself the most powerful video tool out now, at up to 70% off. Cinematic reasoning & Ultimate realism. No more broken physics. Character consistency at it's best. Faster and higher quality than other models on the market. Other platforms don't come near to the convenience of it.

Pretty crazy what a single person can do nowadays by sitting down on their computer. A tool like Cinema Studio 3.0 takes it up a notch, and let's you direct your own films, with the exact characters, locations, and stories you're been aching to craft. Test yourself the most powerful video tool out now, at up to 70% off. Cinematic reasoning & Ultimate realism. No more broken physics. Character consistency at it's best. Faster and higher quality than other models on the market. Other platforms don't come near to the convenience of it.

P̶a̶i̶n̶

13,554 次观看 • 3 个月前

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

kwindla

40,319 次观看 • 1 个月前

0xSero helped me setup local models properly and I uh, had no idea these things had gotten this good Are they frontier level? No, but considering this is running on just my 5090 it's remarkably capable First tests on a couple of programming tasks and the qwen 3.6-27b model with no reasoning feels about on par with something like sonnet 4-ish, probably better it's really impressive But also setting up local models isn't easy, I don't know nearly enough to talk much about it yet other than you need to know what you're doing to have a good experience. The out of the box stuff is not nearly as good as setting it up correctly

0xSero helped me setup local models properly and I uh, had no idea these things had gotten this good Are they frontier level? No, but considering this is running on just my 5090 it's remarkably capable First tests on a couple of programming tasks and the qwen 3.6-27b model with no reasoning feels about on par with something like sonnet 4-ish, probably better it's really impressive But also setting up local models isn't easy, I don't know nearly enough to talk much about it yet other than you need to know what you're doing to have a good experience. The out of the box stuff is not nearly as good as setting it up correctly

Ben Davis

14,490 次观看 • 2 个月前

Superior methodology beats raw parameter count. 🧠 Introducing QVAC MedPsy: Local-first medical AI that redefines the possible. 1/ Unprecedented Power: MedPsy 1.7B model outperforms Google’s MedGemma 4B by 11 points and our 4B model beats MedGemma 27B on real-world health benchmarks. 2/ Extreme Efficiency: 3.2x fewer tokens means near-instant inference on your phone or wearable. 3/ Absolute Privacy: Expert-level reasoning running 100% locally. No data leaves your device. We aren’t simply shrinking models; we’re anchoring intelligence where it matters most. High-level medical logic is now a sovereign right. The future of healthcare is local. Learn more:

Superior methodology beats raw parameter count. 🧠 Introducing QVAC MedPsy: Local-first medical AI that redefines the possible. 1/ Unprecedented Power: MedPsy 1.7B model outperforms Google’s MedGemma 4B by 11 points and our 4B model beats MedGemma 27B on real-world health benchmarks. 2/ Extreme Efficiency: 3.2x fewer tokens means near-instant inference on your phone or wearable. 3/ Absolute Privacy: Expert-level reasoning running 100% locally. No data leaves your device. We aren’t simply shrinking models; we’re anchoring intelligence where it matters most. High-level medical logic is now a sovereign right. The future of healthcare is local. Learn more:

QVAC

2,416,342 次观看 • 2 个月前