Video yükleniyor...

Video Yüklenemedi

Ana Sayfaya Dön

A new model called Space Bunny (stealth model) is free on open router • 1M context • Text, image + video input • Adjustable reasoning • Tool calling OpenCode is also offering a free preview for this week with zero data retention. Try it on your next coding task....

heyshrutimishra's profile picture

Shruti

181,847 subscribers

16,948 görüntüleme • 3 gün önce •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

27 Yorum

Shruti profil fotoğrafı
Shruti3 gün önce

You can try it here:

Muhammad Ayan profil fotoğrafı
Muhammad Ayan3 gün önce

Tried it today, I got a 3D flight tracker from a sketch in 3.5 minutes :)

Khilesh | AI Tools profil fotoğrafı
Khilesh | AI Tools3 gün önce

1M context is a huge advantage for serious coding tasks.

Mishika AI profil fotoğrafı
Mishika AI3 gün önce

Space Bunny being free on OpenRouter is already interesting.

Samuel Hu profil fotoğrafı
Samuel Hu3 gün önce

1m context is wild, but the first thing i'd test is a 200k-token repo with one stale AGENTS.md and a screenshot assertion. fast demos hide where it starts dropping constraints. did you try a long tool-calling run or just the build prompt?

notgwapo profil fotoğrafı
notgwapo3 gün önce

space bunny sounds wild, ngl. lowkey wanna try it out on my next project. 👀

Farhan Azad Shuvra profil fotoğrafı
Farhan Azad Shuvra3 gün önce

That's impressive!

Ulubatlıhasanım. profil fotoğrafı
Ulubatlıhasanım.3 gün önce

Adjustable reasoning is probably the feature I’d experiment with first.

Ryan Carter profil fotoğrafı
Ryan Carter3 gün önce

1M context could be seriously useful when working across large codebases and multiple files.

Re seven profil fotoğrafı
Re seven3 gün önce

Wonderful

Yasir Ai profil fotoğrafı
Yasir Ai3 gün önce

1M context, multimodal input, adjustable reasoning, and tool calling in one free model is a pretty strong package for developers.

Justin Brave💡 profil fotoğrafı
Justin Brave💡3 gün önce

Fantastic share

Daniel Jossr profil fotoğrafı
Daniel Jossr3 gün önce

Text, image, and video input in one model opens up some really interesting workflows.

AI VerseX🚀 profil fotoğrafı
AI VerseX🚀3 gün önce

A stealth model suddenly becoming available for free is a pretty interesting move.

Mr. Jason💡 profil fotoğrafı
Mr. Jason💡3 gün önce

This is the kind of AI thread that provides real practical value.

Holland Tech Marco profil fotoğrafı
Holland Tech Marco3 gün önce

Tools like this show how quickly AI is moving from experimentation into everyday workflows.

Dhruv kumar profil fotoğrafı
Dhruv kumar3 gün önce

Amazing share

Jara profil fotoğrafı
Jara3 gün önce

Wonderful

Mimu | AI Tools & News profil fotoğrafı
Mimu | AI Tools & News3 gün önce

Gave it a shot today turned a quick sketch into a full 3D flight tracker in literally 3.5 minutes.

Ashley Nicole profil fotoğrafı
Ashley Nicole3 gün önce

Awesome thanks

Shruti profil fotoğrafı
Shruti3 gün önce

Glad its helpful

Jeremy Longworth profil fotoğrafı
Jeremy Longworth3 gün önce

Stealth models are becoming a pretty good reminder not to overfit your workflow to a brand name. If a free model can take 1M context, reason, see images/video and call tools, the sensible move is to test it on your actual workload and keep whatever clears the bar.

AI Quanting profil fotoğrafı
AI Quanting3 gün önce

Any guesses on whose it is? Video input plus 1M context feels like a pretty short list.

John Smith profil fotoğrafı
John Smith3 gün önce

Done

Vipul Kumar Kewat profil fotoğrafı
Vipul Kumar Kewat3 gün önce

A free 1M-context model with multimodal input and tool calling is definitely worth testing on a real coding workflow.

frank z993 profil fotoğrafı
frank z9933 gün önce

true no token limit on opencode

Evon profil fotoğrafı
Evon3 gün önce

Awesome post

Benzer Videolar

what the hell did this free stealth model just build 💀 I asked Space Bunny, a Flash-level model, to make Minecraft in one HTML file and left it running it came back with a playable survival game inside my browser I could explore the generated world, break and place blocks, and switch items through the hotbar easily the best result I have got from a coding model on this test Space Bunny is free on OpenCode for the next week prompt below:
2:19

what the hell did this free stealth model just build 💀 I asked Space Bunny, a Flash-level model, to make Minecraft in one HTML file and left it running it came back with a playable survival game inside my browser I could explore the generated world, break and place blocks, and switch items through the hotbar easily the best result I have got from a coding model on this test Space Bunny is free on OpenCode for the next week prompt below:

Israfil

41,131 görüntüleme • 5 gün önce

Hy3, the new 295B MoE model from Tencent Hy, is now free in Nous Portal for the next two weeks! It is focused on cost-effective agentic use, and particularly strong on coding, tool-calling reliability, reasoning, and 256K long-context tasks.
0:24

Hy3, the new 295B MoE model from Tencent Hy, is now free in Nous Portal for the next two weeks! It is focused on cost-effective agentic use, and particularly strong on coding, tool-calling reliability, reasoning, and 256K long-context tasks.

Nous Research

281,555 görüntüleme • 2 ay önce

new Ox Alpha is kind of absurd One prompt, 64,745 output tokens, and no external assets: Ox Alpha generated the code for an entire Three.js dreamcore 3D scene/world in a single shot. AI/ML API ran this demo and they are also giving Ox Alpha free on aimlapi[.]com - 1M context window - Max output: ~131k tokens - Multimodal: text + image + video input → text output - Explicitly positioned as a reasoning model for coding, long-horizon/agentic software engineering, sustained production workloads, and tasks that mix text with visual context
0:22

new Ox Alpha is kind of absurd One prompt, 64,745 output tokens, and no external assets: Ox Alpha generated the code for an entire Three.js dreamcore 3D scene/world in a single shot. AI/ML API ran this demo and they are also giving Ox Alpha free on aimlapi[.]com - 1M context window - Max output: ~131k tokens - Multimodal: text + image + video input → text output - Explicitly positioned as a reasoning model for coding, long-horizon/agentic software engineering, sustained production workloads, and tasks that mix text with visual context

Rohan Paul

13,576 görüntüleme • 1 ay önce

By popular demand, Hy3 from Tencent Hy is free in Nous Portal for an extra week! It is focused on cost-effective agentic use, and is particularly strong on coding, tool-calling reliability, reasoning, and long-context tasks. Try Portal today at
0:24

By popular demand, Hy3 from Tencent Hy is free in Nous Portal for an extra week! It is focused on cost-effective agentic use, and is particularly strong on coding, tool-calling reliability, reasoning, and long-context tasks. Try Portal today at

Nous Research

123,835 görüntüleme • 2 ay önce

A new frontier AI model just dropped on OpenRouter and it’s free. Pony Alpha delivers 200k context, visible reasoning, adjustable thinking depth, and agent-ready tool calling that rivals (and beats) Opus 4.5 in benchmarks. This one’s serious.
9:08

A new frontier AI model just dropped on OpenRouter and it’s free. Pony Alpha delivers 200k context, visible reasoning, adjustable thinking depth, and agent-ready tool calling that rivals (and beats) Opus 4.5 in benchmarks. This one’s serious.

Mervin Praison

23,400 görüntüleme • 7 ay önce

I built a 3D Earth and solar system orbit game with a shipping dashboard using Space Bunny, the new mystery AI model. No idea who’s behind it, but here are my initial takeaways: - Latency: Surprisingly snappy. It doesn’t drag even on longer prompts. - Context: Holds up to 1M input / 524k output. - Multimodal: Smoothly reads images and video directly in the middle of a chat without breaking flow. It’s fast, and can feel flash level. It’s sitting on OpenCode for free this week if you want to try it.
1:15

I built a 3D Earth and solar system orbit game with a shipping dashboard using Space Bunny, the new mystery AI model. No idea who’s behind it, but here are my initial takeaways: - Latency: Surprisingly snappy. It doesn’t drag even on longer prompts. - Context: Holds up to 1M input / 524k output. - Multimodal: Smoothly reads images and video directly in the middle of a chat without breaking flow. It’s fast, and can feel flash level. It’s sitting on OpenCode for free this week if you want to try it.

Veee

57,582 görüntüleme • 4 gün önce

introducing rd-signal-2: a frontier classification model that is 1600x cheaper than GPT 5.6 Sol. free to try in Raindrop, and available via a new API for training/hosting custom classifiers with Zero Data Retention.
0:49

introducing rd-signal-2: a frontier classification model that is 1600x cheaper than GPT 5.6 Sol. free to try in Raindrop, and available via a new API for training/hosting custom classifiers with Zero Data Retention.

ben hylak

162,078 görüntüleme • 1 ay önce

SITUATION EXPLAINED: A stealth model on OpenRouter is beating Fable and Sol on coding, and nobody knows who made it. • Ox Alpha has a 1M token context window, text, image, and video input, free for a week, with capacity for 100 trillion tokens a day • On a DeepSWE subset it hit 80%, against Fable's 65%, GLM-5.3's 62%, and Sol's 52% • The tokenizer is one-for-one identical to GLM's, and a live test asking whether Taiwan is part of China returned a nearly identical answer • This is the 5th anonymous drop in six months, and the previous four were all claimed by Chinese labs: Zhipu, Xiaomi, Ant, and Meituan Theo Jaffee: " There's no way it's a Flash model. If that's a Flash model, then holy shit, the entire United States is cooked. If that's a Flash model, we need to short the stock market now. Not financial advice."
15:43

SITUATION EXPLAINED: A stealth model on OpenRouter is beating Fable and Sol on coding, and nobody knows who made it. • Ox Alpha has a 1M token context window, text, image, and video input, free for a week, with capacity for 100 trillion tokens a day • On a DeepSWE subset it hit 80%, against Fable's 65%, GLM-5.3's 62%, and Sol's 52% • The tokenizer is one-for-one identical to GLM's, and a live test asking whether Taiwan is part of China returned a nearly identical answer • This is the 5th anonymous drop in six months, and the previous four were all claimed by Chinese labs: Zhipu, Xiaomi, Ant, and Meituan Theo Jaffee: " There's no way it's a Flash model. If that's a Flash model, then holy shit, the entire United States is cooked. If that's a Flash model, we need to short the stock market now. Not financial advice."

MTS

382,558 görüntüleme • 1 ay önce

This is Claude Sonnet 4.6: our most capable Sonnet model yet. It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It also features a 1M token context window in beta.
1:29

This is Claude Sonnet 4.6: our most capable Sonnet model yet. It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It also features a 1M token context window in beta.

Claude

7,607,892 görüntüleme • 7 ay önce

Space Bunny Alpha is now on OpenRouter. I made this case myself to explore what it could do, and what stood out to me was the portfolio: a dark, editorial feel, sharp typography, and distinct project worlds like “Chroma Drift” and “Glass Garden,” alongside striking 3D and video scenes. The whole thing feels cohesive and art-directed, not like a generic template. From my own testing and personal impression, Space Bunny Alpha feels Flash-level in terms of responsiveness. OpenRouter also lists: - adjustable reasoning - a 1M-token context window - text, image, and video input This clip doesn’t show a speed test or those inputs in action, so I’m curious to push it further on a live build. Give it a try on OpenRouter:
0:58

Space Bunny Alpha is now on OpenRouter. I made this case myself to explore what it could do, and what stood out to me was the portfolio: a dark, editorial feel, sharp typography, and distinct project worlds like “Chroma Drift” and “Glass Garden,” alongside striking 3D and video scenes. The whole thing feels cohesive and art-directed, not like a generic template. From my own testing and personal impression, Space Bunny Alpha feels Flash-level in terms of responsiveness. OpenRouter also lists: - adjustable reasoning - a 1M-token context window - text, image, and video input This clip doesn’t show a speed test or those inputs in action, so I’m curious to push it further on a live build. Give it a try on OpenRouter:

Kylie

54,771 görüntüleme • 2 gün önce

HERMES AGENT SUPPORTS 300+ MODELS. PICKING THE RIGHT ONE PER TASK IS THE DIFFERENCE BETWEEN $5/MONTH AND $50. STARTING OUT: Claude Sonnet 4.6. official recommendation from Nous Research. "the model this project was built and tested with." strong reasoning. reliable tool calling. mid-range pricing. PREMIUM TIER: Claude Opus 4.8. best coding benchmarks available. self-correcting reasoning. catches its own mistakes. 1M context. use for demanding tasks where quality matters. GPT-5.5. #1 Chatbot Arena. #1 GPQA Diamond reasoning (94.1%). #1 creative writing. 2M context. handles entire codebases in one pass. Grok 4.30. the only frontier model with live X firehose access. real-time social data, breaking news, market sentiment. connects via Grok OAuth. no separate API key. Grok-Composer-2.5-Fast (v0.17.0). Cursor's coding model. 200K context. available through your Grok subscription via OAuth. no extra cost if you already pay for Grok. MID-RANGE TIER: Claude Sonnet 4.6. best balance of quality and cost for daily use. strongest prose and tool calling in this tier. Gemini 2.5 Pro. Google Search grounding built in. cites sources. verifies claims. pulls current data. 2M context. best for research-heavy workflows. GPT-4.1. reliable tool calling. solid general reasoning. good middle ground when you need OpenAI compatibility. BUDGET TIER: Claude Haiku 4.5. fastest Anthropic model. cheapest paid Claude option. strong at classification, routing, simple queries. use for auxiliary tasks: compression, vision, web extraction, approval scoring. DeepSeek V4. best cost-to-quality ratio in the market. 90% cache discount on repeated context. use for sub-agents and bulk parallel work. DeepSeek V4 Flash. cheapest paid model worth using. 1M context. MIT license. self-hostable. use for cron jobs, monitoring, routine searches. MiniMax M3. Nous Research and MiniMax collaborating on optimization. 1M context via lightning attention. 59% SWE-Bench Pro. beats several premium models on coding. one of the most-used models inside Hermes. FREE / LOCAL: Qwen 3.5 27B via Ollama. 16GB VRAM. reliable tool calling. best free local model for Hermes as of mid-2026. Qwen 3 8B. 8GB VRAM. fits a $7 VPS. handles routine tasks at zero API cost. Llama 4 Maverick. best open-weight tool calling. 1M context. needs more VRAM but strongest local option. HOW TO ASSIGN MODELS: main model: Desktop app / Dashboard → Models → switch sub-agent model: set in Desktop app, Dashboard, or config.yaml: delegation: model: "deepseek/deepseek-v4" auxiliary models (compression, vision, web extract): Desktop app / Dashboard → Models → Auxiliary Haiku 4.5 or Gemini Flash work well here. saves significantly when your main model is premium. per-profile: each Hermes profile gets its own model. Scout on DeepSeek. Analyst on Sonnet. Briefer on budget model. Coder on Opus. per-cron-job: pin a specific model to any cron job. morning brief on Haiku. deep research on Sonnet. monitoring on DeepSeek Flash. each job uses only the model it needs. per-session: /model deepseek/deepseek-v4-flash hot-swap mid-conversation. no restart needed. FALLBACK CHAINS: if your primary model is unavailable, Hermes automatically switches to the next provider. rate limit or server error = next model in the chain. no failed runs. no manual intervention. set in Desktop app, Dashboard, or config.yaml: fallback_providers: - openrouter - nous - codex PROVIDER PATHS: OPENROUTER: 300+ models under one API key. pay per token. most flexible. NOUS PORTAL: 300+ models + Tool Gateway (web search, image gen, TTS, browser). one OAuth. one subscription. 10% off token-billed providers. CHATGPT SUB: GPT-5.5 + Grok via OAuth. included tokens with $20 subscription. OLLAMA: free. local. private. zero API cost. your hardware only. mix providers across profiles and tasks. Scout on OpenRouter. Analyst on Nous Portal. Coder on ChatGPT sub. Monitor on Ollama. THE RULE: premium for work that needs deep reasoning. mid-range for daily driver tasks. budget for volume and background work. free for monitoring and routine jobs. pricing changes fast. check openrouter ai for current rates before committing. Which is your favourite model and for what task? full 15 levels breakdown in the article 👇
0:51

HERMES AGENT SUPPORTS 300+ MODELS. PICKING THE RIGHT ONE PER TASK IS THE DIFFERENCE BETWEEN $5/MONTH AND $50. STARTING OUT: Claude Sonnet 4.6. official recommendation from Nous Research. "the model this project was built and tested with." strong reasoning. reliable tool calling. mid-range pricing. PREMIUM TIER: Claude Opus 4.8. best coding benchmarks available. self-correcting reasoning. catches its own mistakes. 1M context. use for demanding tasks where quality matters. GPT-5.5. #1 Chatbot Arena. #1 GPQA Diamond reasoning (94.1%). #1 creative writing. 2M context. handles entire codebases in one pass. Grok 4.30. the only frontier model with live X firehose access. real-time social data, breaking news, market sentiment. connects via Grok OAuth. no separate API key. Grok-Composer-2.5-Fast (v0.17.0). Cursor's coding model. 200K context. available through your Grok subscription via OAuth. no extra cost if you already pay for Grok. MID-RANGE TIER: Claude Sonnet 4.6. best balance of quality and cost for daily use. strongest prose and tool calling in this tier. Gemini 2.5 Pro. Google Search grounding built in. cites sources. verifies claims. pulls current data. 2M context. best for research-heavy workflows. GPT-4.1. reliable tool calling. solid general reasoning. good middle ground when you need OpenAI compatibility. BUDGET TIER: Claude Haiku 4.5. fastest Anthropic model. cheapest paid Claude option. strong at classification, routing, simple queries. use for auxiliary tasks: compression, vision, web extraction, approval scoring. DeepSeek V4. best cost-to-quality ratio in the market. 90% cache discount on repeated context. use for sub-agents and bulk parallel work. DeepSeek V4 Flash. cheapest paid model worth using. 1M context. MIT license. self-hostable. use for cron jobs, monitoring, routine searches. MiniMax M3. Nous Research and MiniMax collaborating on optimization. 1M context via lightning attention. 59% SWE-Bench Pro. beats several premium models on coding. one of the most-used models inside Hermes. FREE / LOCAL: Qwen 3.5 27B via Ollama. 16GB VRAM. reliable tool calling. best free local model for Hermes as of mid-2026. Qwen 3 8B. 8GB VRAM. fits a $7 VPS. handles routine tasks at zero API cost. Llama 4 Maverick. best open-weight tool calling. 1M context. needs more VRAM but strongest local option. HOW TO ASSIGN MODELS: main model: Desktop app / Dashboard → Models → switch sub-agent model: set in Desktop app, Dashboard, or config.yaml: delegation: model: "deepseek/deepseek-v4" auxiliary models (compression, vision, web extract): Desktop app / Dashboard → Models → Auxiliary Haiku 4.5 or Gemini Flash work well here. saves significantly when your main model is premium. per-profile: each Hermes profile gets its own model. Scout on DeepSeek. Analyst on Sonnet. Briefer on budget model. Coder on Opus. per-cron-job: pin a specific model to any cron job. morning brief on Haiku. deep research on Sonnet. monitoring on DeepSeek Flash. each job uses only the model it needs. per-session: /model deepseek/deepseek-v4-flash hot-swap mid-conversation. no restart needed. FALLBACK CHAINS: if your primary model is unavailable, Hermes automatically switches to the next provider. rate limit or server error = next model in the chain. no failed runs. no manual intervention. set in Desktop app, Dashboard, or config.yaml: fallback_providers: - openrouter - nous - codex PROVIDER PATHS: OPENROUTER: 300+ models under one API key. pay per token. most flexible. NOUS PORTAL: 300+ models + Tool Gateway (web search, image gen, TTS, browser). one OAuth. one subscription. 10% off token-billed providers. CHATGPT SUB: GPT-5.5 + Grok via OAuth. included tokens with $20 subscription. OLLAMA: free. local. private. zero API cost. your hardware only. mix providers across profiles and tasks. Scout on OpenRouter. Analyst on Nous Portal. Coder on ChatGPT sub. Monitor on Ollama. THE RULE: premium for work that needs deep reasoning. mid-range for daily driver tasks. budget for volume and background work. free for monitoring and routine jobs. pricing changes fast. check openrouter ai for current rates before committing. Which is your favourite model and for what task? full 15 levels breakdown in the article 👇

YanXbt

17,138 görüntüleme • 3 ay önce

Meta Tribe2 is a model that predicts your brain's response on videos so now you can edit videos with more clarity of how to compose them, what works and what doesn't and we built a free tool for you based on this model yes, totally free see details below
0:24

Meta Tribe2 is a model that predicts your brain's response on videos so now you can edit videos with more clarity of how to compose them, what works and what doesn't and we built a free tool for you based on this model yes, totally free see details below

AmirMušić

853,555 görüntüleme • 5 ay önce

OpenAI o1 is now out of preview in ChatGPT. What’s changed since the preview? A faster, more powerful reasoning model that’s better at coding, math & writing. o1 now also supports image uploads, allowing it to apply reasoning to visuals for more detailed & useful responses.
0:35

OpenAI o1 is now out of preview in ChatGPT. What’s changed since the preview? A faster, more powerful reasoning model that’s better at coding, math & writing. o1 now also supports image uploads, allowing it to apply reasoning to visuals for more detailed & useful responses.

OpenAI

2,344,637 görüntüleme • 1 yıl önce

Maybe I'm smoking something, but this is what I got out of a Qwen3.8-27b on Agent Mode. Alibaba's endpoint is running at 1m context, and this was about 260k output tokens. I don't remember this from a 27b model. I also did a video with more tests, link below
0:40

Maybe I'm smoking something, but this is what I got out of a Qwen3.8-27b on Agent Mode. Alibaba's endpoint is running at 1m context, and this was about 260k output tokens. I don't remember this from a 27b model. I also did a video with more tests, link below

Peter Gostev (SF 24-28 August)

33,112 görüntüleme • 1 ay önce

Exactly AI Illustrative Training is now on Runware⚙️ - Train a custom model on your brand's visual style - 10–50 images produces a reusable brand-aligned model - Text-to-image + image-to-image with zero style prompting required - Consistent output across campaigns & products
0:21

Exactly AI Illustrative Training is now on Runware⚙️ - Train a custom model on your brand's visual style - 10–50 images produces a reusable brand-aligned model - Text-to-image + image-to-image with zero style prompting required - Consistent output across campaigns & products

Runware

578,364 görüntüleme • 4 ay önce

Wait Xiaomi has also released the best 9B model in the world?! This model is based on Qwen-9B... And you can easily run it even on 8GB of memory! - MIT license - Multimodal (image + text) - Solid for coding/agentic tasks Perfect to use locally and offline... even on a laptop with unified memory!
0:37

Wait Xiaomi has also released the best 9B model in the world?! This model is based on Qwen-9B... And you can easily run it even on 8GB of memory! - MIT license - Multimodal (image + text) - Solid for coding/agentic tasks Perfect to use locally and offline... even on a laptop with unified memory!

Paul Couvert

84,556 görüntüleme • 5 gün önce

Gemma 4 31B is now available in Public Preview on Cerebras. Our first multimodal model runs at over 1,800 tokens/s for ultra-fast image and text workflows. Give it a try:
2:08

Gemma 4 31B is now available in Public Preview on Cerebras. Our first multimodal model runs at over 1,800 tokens/s for ultra-fast image and text workflows. Give it a try:

Cerebras

282,043 görüntüleme • 3 ay önce

Qwopus on a single RTX 3090. Claude Opus 4.6 reasoning distilled into Qwen 3.5 27B dense, running through Claude's own coding agent (claude code). 29-35 tok/s with thinking mode on. the jinja bug that kills thinking on base Qwen doesn't carry over. harness and model matched. the base model would pause mid task on Claude Code. just stop generating. that's why i ran it through OpenCode, which handles stalled states automatically. this distilled version doesn't stall. it waits for tool outputs, reads them, selfcorrects when something breaks, and keeps going. i gave it a benchmark analysis task. went 9 minutes autonomous. wrote a README nobody asked for. zero steering. video is 5x speed but fully uncut. if you have a 3090, you can run this right now. free. no API. no subscription. opus structured reasoning on localhost. octopus invaders is next. same prompt that base qwen passed in 13 minutes and hermes 4.3 failed on 2x the hardware. i want to see if the distillation changes the outcome or just the style. more data soon.
1:53

Qwopus on a single RTX 3090. Claude Opus 4.6 reasoning distilled into Qwen 3.5 27B dense, running through Claude's own coding agent (claude code). 29-35 tok/s with thinking mode on. the jinja bug that kills thinking on base Qwen doesn't carry over. harness and model matched. the base model would pause mid task on Claude Code. just stop generating. that's why i ran it through OpenCode, which handles stalled states automatically. this distilled version doesn't stall. it waits for tool outputs, reads them, selfcorrects when something breaks, and keeps going. i gave it a benchmark analysis task. went 9 minutes autonomous. wrote a README nobody asked for. zero steering. video is 5x speed but fully uncut. if you have a 3090, you can run this right now. free. no API. no subscription. opus structured reasoning on localhost. octopus invaders is next. same prompt that base qwen passed in 13 minutes and hermes 4.3 failed on 2x the hardware. i want to see if the distillation changes the outcome or just the style. more data soon.

Sudo su

296,392 görüntüleme • 6 ay önce

🎉 Congrats to MiniMax (official) on releasing MiniMax M3! Frontier coding and agentic capabilities, native image and video input, computer use, and a 1M-token context window, all in a single open model. At the heart of M3 is MSA, a new sparse attention architecture: instead of attending densely over the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗
0:21

🎉 Congrats to MiniMax (official) on releasing MiniMax M3! Frontier coding and agentic capabilities, native image and video input, computer use, and a 1M-token context window, all in a single open model. At the heart of M3 is MSA, a new sparse attention architecture: instead of attending densely over the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗

vLLM

40,306 görüntüleme • 3 ay önce