Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Sam Altman: the ideal AI is a tiny model with superhuman reasoning, a trillion-token context window, and access to every tool. “The platonic ideal is a very tiny model that has superhuman reasoning capabilities.” “It can run ridiculously fast.” “One trillion tokens of context and access to every tool... show more

Veee

9,468 subscribers

68,160 Aufrufe • vor 7 Tagen •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

7 Kommentare

Profilbild von SwingTraderStocks

SwingTraderStocksvor 7 Tagen

🆘“Using these models as databases is sort of ridiculous, ...very broken databases... but the amazing thing is they can read them"⛽️

Profilbild von CJ

CJvor 7 Tagen

We’re headed there, just not with your models.

Profilbild von Mehdi Mohseni

Mehdi Mohsenivor 7 Tagen

Access to every tool is the part I would push back on. Past a certain count the model stops picking well, so you end up building discovery and scoping anyway. A small, well described tool surface beats a huge one, whatever the context size.

Profilbild von foombler 🐂🃏

foombler 🐂🃏vor 7 Tagen

this ideal model is also the best fit to be deployed into edge hardware like robotics platforms with compute constraints

Profilbild von GBE

GBEvor 7 Tagen

new model up next lets gooo vee

Profilbild von Laff

Laffvor 7 Tagen

Bold claim.

Profilbild von Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿

Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿vor 6 Tagen

The ideal AI is a specialized tiny model with communication to other specialized models.

Ähnliche Videos

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.

Sam Altman says the perfect AI is “a very tiny model with superhuman reasoning, 1 trillion tokens of context, and access to every tool you can imagine.” It doesn't need to contain the knowledge - just the ability to think, search, simulate, and solve anything.

vitrupo

775,853 Aufrufe • vor 1 Jahr

Qwen3.8-Flash-Next is still going strong at 364.7K tokens of context on an M5 Max. And this isn’t just a static long-context test. The model was reasoning about how to speed up its own workflow while using tools, and the tool calls kept working without misses. Setup: • Qwen3.8-Flash-Next • M5 Max • 128GB unified memory • MLX-Serve PR #363 • OpenCode 2 • 364.7K context The interesting part isn’t simply getting hundreds of thousands of tokens into memory. It’s what happens once the context gets this large. Long-context inference usually comes with a painful tradeoff. As the KV cache grows, memory pressure increases and generation can slow down. But this setup is still pushing through 364K tokens while maintaining a usable agent workflow. The model can reason, call tools, inspect results, continue working, and keep the session moving. And the tool calls reportedly haven’t missed so far. That’s important for agentic coding. A huge context window is only useful if the model can actually operate reliably inside it. A 400K-token context that constantly breaks tool calls isn’t very useful. A 364K session that can keep reasoning and executing tools is a different story. And the test isn’t finished yet. The current run is approaching 400K tokens, with the expectation that it can keep going. This is also another interesting example of why Apple Silicon keeps showing up in local LLM experiments. The M5 Max’s unified memory gives a large model and its growing KV cache access to one shared memory pool. With MLX-Serve continuing to improve, these machines are becoming surprisingly capable long-context inference boxes. The bigger takeaway: Context length is becoming a workload, not just a model specification. Running a model at 256K is one thing. Keeping an agent alive at 300K+ while it reasons and uses tools is much more interesting. And Qwen3.8-Flash-Next is showing that this can be pushed surprisingly far on a single 128GB Mac. 364.7K and counting. Next stop: 400K.

Qwen3.8-Flash-Next is still going strong at 364.7K tokens of context on an M5 Max. And this isn’t just a static long-context test. The model was reasoning about how to speed up its own workflow while using tools, and the tool calls kept working without misses. Setup: • Qwen3.8-Flash-Next • M5 Max • 128GB unified memory • MLX-Serve PR #363 • OpenCode 2 • 364.7K context The interesting part isn’t simply getting hundreds of thousands of tokens into memory. It’s what happens once the context gets this large. Long-context inference usually comes with a painful tradeoff. As the KV cache grows, memory pressure increases and generation can slow down. But this setup is still pushing through 364K tokens while maintaining a usable agent workflow. The model can reason, call tools, inspect results, continue working, and keep the session moving. And the tool calls reportedly haven’t missed so far. That’s important for agentic coding. A huge context window is only useful if the model can actually operate reliably inside it. A 400K-token context that constantly breaks tool calls isn’t very useful. A 364K session that can keep reasoning and executing tools is a different story. And the test isn’t finished yet. The current run is approaching 400K tokens, with the expectation that it can keep going. This is also another interesting example of why Apple Silicon keeps showing up in local LLM experiments. The M5 Max’s unified memory gives a large model and its growing KV cache access to one shared memory pool. With MLX-Serve continuing to improve, these machines are becoming surprisingly capable long-context inference boxes. The bigger takeaway: Context length is becoming a workload, not just a model specification. Running a model at 256K is one thing. Keeping an agent alive at 300K+ while it reasons and uses tools is much more interesting. And Qwen3.8-Flash-Next is showing that this can be pushed surprisingly far on a single 128GB Mac. 364.7K and counting. Next stop: 400K.

FHILY👑

39,982 Aufrufe • vor 21 Tagen

This is very late, but I'm finally done with my 5.5 vid - use low reasoning - the name sucks - it's fast - best code I've ever seen a model write came from this model - openai's new pre-training is amazing - price looks worse than it is - over sensitive to every little thing in it's context window - feels wildly different compared to 5.4 - turn reasoning off. try it. turn the reasoning off. do it.

This is very late, but I'm finally done with my 5.5 vid - use low reasoning - the name sucks - it's fast - best code I've ever seen a model write came from this model - openai's new pre-training is amazing - price looks worse than it is - over sensitive to every little thing in it's context window - feels wildly different compared to 5.4 - turn reasoning off. try it. turn the reasoning off. do it.

Ben Davis

283,972 Aufrufe • vor 4 Monaten

“A tiny reasoning model, trillions of tokens, that has heard every conversation you’ve ever had” Are Sam and Jony Ive building an omnipresent wearable listening and vision device? Here’s Altman on Building the ‘Core AI Subscription’ for Your Life:

“A tiny reasoning model, trillions of tokens, that has heard every conversation you’ve ever had” Are Sam and Jony Ive building an omnipresent wearable listening and vision device? Here’s Altman on Building the ‘Core AI Subscription’ for Your Life:

AI Breakfast

163,241 Aufrufe • vor 1 Jahr

The MiniMax M2 model is mind-blowing! It's open-source. It outperforms Gemini 2.5, Claude 4.1, and Qwen3 across coding and tool-use benchmarks. Right now, it's one of the world's top 5 models in intelligence! And here is the best part: Claude is one of the best models you can use today, and MiniMax M2 costs only 8% of that! It's smaller, faster, and cheaper. Extremely efficient at using tokens. Minimax M2's biggest strength: High agentic capabilities. The model can plan and execute complex multi-tool workflows. It's reliable and very robust at executing long-horizon tool chains. In summary: • Low latency • Very cheap • Excels at agentic tasks • Open-source The model currently powers the MiniMax Agent and is available for a free global trial. You can access MiniMax M2's API here: To access the agent: And here is the MiniMax website: Thanks to the MiniMax team for showing me the ropes and partnering with me on this post.

The MiniMax M2 model is mind-blowing! It's open-source. It outperforms Gemini 2.5, Claude 4.1, and Qwen3 across coding and tool-use benchmarks. Right now, it's one of the world's top 5 models in intelligence! And here is the best part: Claude is one of the best models you can use today, and MiniMax M2 costs only 8% of that! It's smaller, faster, and cheaper. Extremely efficient at using tokens. Minimax M2's biggest strength: High agentic capabilities. The model can plan and execute complex multi-tool workflows. It's reliable and very robust at executing long-horizon tool chains. In summary: • Low latency • Very cheap • Excels at agentic tasks • Open-source The model currently powers the MiniMax Agent and is available for a free global trial. You can access MiniMax M2's API here: To access the agent: And here is the MiniMax website: Thanks to the MiniMax team for showing me the ropes and partnering with me on this post.

Santiago

91,240 Aufrufe • vor 11 Monaten

Our latest speech-to-speech model is faster, more accurate, and excels at function calling. Watch @promptshant and Brian Fioca build a realtime voice agent that can search the web and hand off tasks to reasoning models with full context.

Our latest speech-to-speech model is faster, more accurate, and excels at function calling. Watch @promptshant and Brian Fioca build a realtime voice agent that can search the web and hand off tasks to reasoning models with full context.

OpenAI Developers

81,863 Aufrufe • vor 1 Jahr

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

40,593 Aufrufe • vor 4 Monaten

Sam Altman says AI reasoning is still at the GPT-2 stage but the improvement curve is steep and the new o1 model represents a new paradigm of AI development which will enable rapid progress in capabilities

Sam Altman says AI reasoning is still at the GPT-2 stage but the improvement curve is steep and the new o1 model represents a new paradigm of AI development which will enable rapid progress in capabilities

Tsarathustra

419,003 Aufrufe • vor 2 Jahren

they gave frontier models a 1 billion token context window: you can now run Fable [1B], Opus [100M], GPT 5.6 [500M] across your entire codebase - it's a graph based tool that can traverse databases with 400M+ of nodes in ms - only one prompt to install - & it's open source (link in video) Introducing Polygres

they gave frontier models a 1 billion token context window: you can now run Fable [1B], Opus [100M], GPT 5.6 [500M] across your entire codebase - it's a graph based tool that can traverse databases with 400M+ of nodes in ms - only one prompt to install - & it's open source (link in video) Introducing Polygres

Founders Inc

400,430 Aufrufe • vor 1 Monat

As one example of 1.5 Pro’s sophisticated multimodal understanding and reasoning capabilities with long context, when given a 44-minute silent film, the model can analyze various plot points and events, and even makes sense of small details you might have missed.

As one example of 1.5 Pro’s sophisticated multimodal understanding and reasoning capabilities with long context, when given a 44-minute silent film, the model can analyze various plot points and events, and even makes sense of small details you might have missed.

Sundar Pichai

348,411 Aufrufe • vor 2 Jahren

Sam Altman says the day is approaching when we can ask an AI model to solve all of physics and it can actually do that

Sam Altman says the day is approaching when we can ask an AI model to solve all of physics and it can actually do that

Tsarathustra

852,780 Aufrufe • vor 2 Jahren

NEW: LiquidAI just released LFM2.5-350M, a tiny model that brings agentic AI and tool-calling capabilities to resource-constrained environments. 🤯 It can even run locally in your browser via WebGPU, serving as a powerful companion while you browse the web. Try the demo! 👇

NEW: LiquidAI just released LFM2.5-350M, a tiny model that brings agentic AI and tool-calling capabilities to resource-constrained environments. 🤯 It can even run locally in your browser via WebGPU, serving as a powerful companion while you browse the web. Try the demo! 👇

Xenova

26,267 Aufrufe • vor 6 Monaten

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.

How can you solve complex tasks using a Large Language Model? Here is a 2-minute introduction to everything you need to know to 10x the quality of your results. Let's talk about three techniques, in order of complexity, starting with the easiest one: • In-Context Learning • Indexing + In-Context Learning • Fine-tuning In-Context Learning The team that trained GPT-3 found something they couldn't explain: You can condition a model using examples of how you want it to behave. I included an example prompt in the attached video. You can "teach" the model how you want it to interpret questions, select the correct answers, and format the results by giving a few examples. You can also give specific knowledge to the model that will be helpful when formulating answers. We call this approach "grounding the model." There's another example in the video. Indexing + In-Context Learning Unfortunately, there is a limit to how much data you can include in a prompt. We call this the "context size." One version of GPT-4 supports a context of approximately 6,000 words, while the other supports 25,000 words. Although this sounds like a lot, many applications need more than that. Imagine you wrote a book and want to build an application to answer any questions about your story. What happens if your book is longer than the context? That's where Indexing comes in. Using a model, you can turn every book passage into an embedding. These are vectors, numbers that "encode" the passage's text. You can then store these embeddings in a particular database that supports fast retrieval of these vectors. You can then turn any question into an embedding and search the database for the list of passages that are similar to that query. Instead of using the entire book to ask the model, you can now use the relevant passages as in-context information, effectively working around the context size limitation. Fine-tuning Fine-tuning can give you an extra boost to get reliable outputs from your LLM. It is, however, the most complex approach on the list. There are different approaches to fine-tuning a model with your data. A popular technique is to process your data with your LLM and use the outputs to train a new classifier that solves your specific task. Notice that here you aren't modifying the LLM. Instead, you are chaining it with your trained classifier. Another approach is to modify the parameters of the LLM using your data. Think of this as "rewiring" the model in a way that solves your particular task. The results and costs will vary depending on how many layers you want to fine-tune from the original model. Many companies think that fine-tuning is the solution to their problems. In my experience, many will benefit from exploring the other two approaches. I love explaining Machine Learning and Artificial Intelligence ideas. If you enjoy in-depth content like this, follow me Santiago so you don't miss what comes next.

Santiago

384,573 Aufrufe • vor 3 Jahren

DeepMind's Nikolay Savinov says 10M-token context windows will transform how AI works. AI will ingest entire codebases at once, becoming "totally unrivaled… the new tool for every coder in the world." 100M is coming too -- and with it, reasoning across systems we can't yet imagine.

DeepMind's Nikolay Savinov says 10M-token context windows will transform how AI works. AI will ingest entire codebases at once, becoming "totally unrivaled… the new tool for every coder in the world." 100M is coming too -- and with it, reasoning across systems we can't yet imagine.

vitrupo

261,728 Aufrufe • vor 1 Jahr

This is the fastest I've seen Llama 3.3 running anywhere! Llama 3.3 70B running at 652 t/s is lightning fast. And if you want Llama 3.1, here are the speeds I was able to get: • Llama 3.1 8B: 1006 t/s • Llama 3.1 70B: 709 t/s • Llama 3.1 405B: 206 t/s (You can access all of these models for free! See the link below.) This speed is incredible, but the interesting part is what's happening behind the scenes: These models aren't running on a GPU! In this video, I'm using the SambaNova cloud to access these models. They built a custom chip (SN40L) optimized for AI workflows. A single SN40L chip can hold hundreds of models (trillions of parameters) in memory! The speed alone is a huge deal, but the big advantage is for agentic workflows running multiple specialized models. A GPU can only host a single model and switch (unload and load) to a different model if necessary. An SN40L, on the other hand, can host every model at once, making it much faster. Here is the video where you can see how fast these chips are:

This is the fastest I've seen Llama 3.3 running anywhere! Llama 3.3 70B running at 652 t/s is lightning fast. And if you want Llama 3.1, here are the speeds I was able to get: • Llama 3.1 8B: 1006 t/s • Llama 3.1 70B: 709 t/s • Llama 3.1 405B: 206 t/s (You can access all of these models for free! See the link below.) This speed is incredible, but the interesting part is what's happening behind the scenes: These models aren't running on a GPU! In this video, I'm using the SambaNova cloud to access these models. They built a custom chip (SN40L) optimized for AI workflows. A single SN40L chip can hold hundreds of models (trillions of parameters) in memory! The speed alone is a huge deal, but the big advantage is for agentic workflows running multiple specialized models. A GPU can only host a single model and switch (unload and load) to a different model if necessary. An SN40L, on the other hand, can host every model at once, making it much faster. Here is the video where you can see how fast these chips are:

Santiago

97,561 Aufrufe • vor 1 Jahr

Glen is building the organizational learning and context layer for agents. Every agent session your team runs produces reasoning worth keeping, and nearly all of it is discarded the moment the window closes. Glen is the layer that unifies every agent and model your company run. Glen pulls in agent sessions from Claude Code, Codex and Cursor, plus Slack messages, GitHub prs, tickets, docs and calls. When a new agent picks up a task, Glen finds relevant prior context and injects it in the agents context window to steer it, producing knowledge work that is significantly higher quality, faster and cheaper than ever before. 29% less context and 21% faster to be exact. And Glen does more... > query the original agent behind any line of code > search every session your org has ever run > dynamic skill distillation > multiplayer agents > unified transcripts you can hand between Claude Code, Codex, and Cursor mid-task, > a company brain in Slack. Glen is in early access, sign up for the waitlist below.

Glen is building the organizational learning and context layer for agents. Every agent session your team runs produces reasoning worth keeping, and nearly all of it is discarded the moment the window closes. Glen is the layer that unifies every agent and model your company run. Glen pulls in agent sessions from Claude Code, Codex and Cursor, plus Slack messages, GitHub prs, tickets, docs and calls. When a new agent picks up a task, Glen finds relevant prior context and injects it in the agents context window to steer it, producing knowledge work that is significantly higher quality, faster and cheaper than ever before. 29% less context and 21% faster to be exact. And Glen does more... > query the original agent behind any line of code > search every session your org has ever run > dynamic skill distillation > multiplayer agents > unified transcripts you can hand between Claude Code, Codex, and Cursor mid-task, > a company brain in Slack. Glen is in early access, sign up for the waitlist below.

Nikos

19,883 Aufrufe • vor 1 Monat

Today we entered the Gemini 3 era, our next step on the path toward AGI. ⚡ Gemini 3 is our most intelligent model that combines capabilities like multimodality, long context and reasoning, so you can bring any idea to life. Explore more of what you can do and build with Gemini 3 🧵⬇️

Today we entered the Gemini 3 era, our next step on the path toward AGI. ⚡ Gemini 3 is our most intelligent model that combines capabilities like multimodality, long context and reasoning, so you can bring any idea to life. Explore more of what you can do and build with Gemini 3 🧵⬇️

Google

90,913 Aufrufe • vor 10 Monaten

Qwen3.8-27B running at full BF16 on a free Kaggle TPU is kind of ridiculous. No quantization. No tiny context window. No expensive GPU instance. Just Qwen3.8-27B running on a Kaggle TPU v5e-8. The reported numbers: ~130 tok/s decode ~10,000 tok/s prefill 262K context That prefill number is especially wild. You can throw a huge amount of code or context at the model and ingest it extremely quickly, while still getting around 130 tokens per second during generation. And because it’s running in full BF16, you’re not relying on an aggressive quant just to make the model fit. But the really interesting part isn’t even the raw throughput. You can expose it as an OpenAI-compatible endpoint. That means you can plug the model into tools that already understand OpenAI-style APIs. Claude Code. Codex. OpenCode. And other compatible clients. So the workflow becomes pretty simple: Spin up the Qwen3.8-27B endpoint on Kaggle. Point your coding tool at the API. And suddenly you have a 27B coding model sitting behind the same interface you’d normally use for hosted models. The 262K context is also a huge deal for agentic coding. Large repositories can fit into a single context. Long conversations don’t need to be constantly trimmed. And tools can feed much more information back to the model without hitting a tiny context ceiling. The fact that this can be built around a free TPU environment is what makes this especially interesting. We’re getting to a point where experimenting with serious open models doesn’t always require owning a $2,000 GPU or paying for a large cloud instance. Free compute + open weights + an OpenAI-compatible API + existing coding agents. That’s a pretty powerful combination. Qwen3.8-27B is already an interesting model. Running the full BF16 version at ~130 tok/s with 262K context on free Kaggle TPU compute makes it a lot more interesting.

Qwen3.8-27B running at full BF16 on a free Kaggle TPU is kind of ridiculous. No quantization. No tiny context window. No expensive GPU instance. Just Qwen3.8-27B running on a Kaggle TPU v5e-8. The reported numbers: ~130 tok/s decode ~10,000 tok/s prefill 262K context That prefill number is especially wild. You can throw a huge amount of code or context at the model and ingest it extremely quickly, while still getting around 130 tokens per second during generation. And because it’s running in full BF16, you’re not relying on an aggressive quant just to make the model fit. But the really interesting part isn’t even the raw throughput. You can expose it as an OpenAI-compatible endpoint. That means you can plug the model into tools that already understand OpenAI-style APIs. Claude Code. Codex. OpenCode. And other compatible clients. So the workflow becomes pretty simple: Spin up the Qwen3.8-27B endpoint on Kaggle. Point your coding tool at the API. And suddenly you have a 27B coding model sitting behind the same interface you’d normally use for hosted models. The 262K context is also a huge deal for agentic coding. Large repositories can fit into a single context. Long conversations don’t need to be constantly trimmed. And tools can feed much more information back to the model without hitting a tiny context ceiling. The fact that this can be built around a free TPU environment is what makes this especially interesting. We’re getting to a point where experimenting with serious open models doesn’t always require owning a $2,000 GPU or paying for a large cloud instance. Free compute + open weights + an OpenAI-compatible API + existing coding agents. That’s a pretty powerful combination. Qwen3.8-27B is already an interesting model. Running the full BF16 version at ~130 tok/s with 262K context on free Kaggle TPU compute makes it a lot more interesting.

FHILY👑

35,910 Aufrufe • vor 22 Tagen