Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

I'm still amazed that Apple Neural Engine can now code on M6 Mini. Qwen 27B really needs reasoning ON for coding. Here is an additional profile for Droid agent w/reasoning.

Anemll

5,400 subscribers

17,042 görüntüleme • 8 gün önce •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

6 Yorum

Liu Liu profil fotoğrafı

Liu Liu8 gün önce

@droid Yeah, reasoning and 0.9 temp.

Dan Woods profil fotoğrafı

Dan Woods8 gün önce

Pretty sick!

AI Quanting profil fotoğrafı

AI Quanting8 gün önce

@droid What's tok/s with reasoning on? That's where on-device coding usually falls apart.

Anemll profil fotoğrafı

Anemll8 gün önce

@droid reasoning tokens don't like dflash2, ~16-20 is usual. I'm working on improvements for that

AI Quanting profil fotoğrafı

AI Quanting8 gün önce

@droid Good to know, thanks. Is the fix on the attention side or the sampler?

La Voix du Chat Artiste profil fotoğrafı

La Voix du Chat Artiste8 gün önce

@droid Should I stick to m6 32 Go instead of MacStudio 36 Go , then ? Real question ! Because I still can’t decide , that is why I haven’t any for the moment Please help !!

Benzer Videolar

The new Qwen 3.5 by Qwen running on-device on iPhone 17 Pro. Qwen 3.5 beats models 4 times its size, has strong visual understanding, and can toggle reasoning on or off. The 2B 6-bit model here is running with MLX optimized for Apple Silicon.

The new Qwen 3.5 by Qwen running on-device on iPhone 17 Pro. Qwen 3.5 beats models 4 times its size, has strong visual understanding, and can toggle reasoning on or off. The 2B 6-bit model here is running with MLX optimized for Apple Silicon.

Adrien Grondin

3,560,817 görüntüleme • 7 ay önce

AI Coding Agent for Hardware Optimized Code Diana AI hardware is still constrained by software. However, with reasoning models like Deepseek R1 or OpenAI o1 and o3, AI could generate hardware-optimized code that rivals—or surpasses—human CUDA code.

AI Coding Agent for Hardware Optimized Code Diana AI hardware is still constrained by software. However, with reasoning models like Deepseek R1 or OpenAI o1 and o3, AI could generate hardware-optimized code that rivals—or surpasses—human CUDA code.

Y Combinator

60,185 görüntüleme • 1 yıl önce

I built a Deepseek R1 RAG Reasoning Agent running locally on my computer. It's an Agentic RAG reasoning agent that can think, reason and fall back to web search if needed. 100% Opensource code with step-by-step tutorial.

I built a Deepseek R1 RAG Reasoning Agent running locally on my computer. It's an Agentic RAG reasoning agent that can think, reason and fall back to web search if needed. 100% Opensource code with step-by-step tutorial.

Shubham Saboo

124,490 görüntüleme • 1 yıl önce

60 FPS body tracking + mask segmentation at 2K in portrait, off GPU! Running on Apple Neural Engine: Body pose:~9ms (6x faster than Mediapipe) Mask:~9ms (quality is super impressive) GPU is still free for Three.js fun! My goal is advanced installation for <1000$ (M6 Mac Mini)

60 FPS body tracking + mask segmentation at 2K in portrait, off GPU! Running on Apple Neural Engine: Body pose:~9ms (6x faster than Mediapipe) Mask:~9ms (quality is super impressive) GPU is still free for Three.js fun! My goal is advanced installation for <1000$ (M6 Mac Mini)

Renaud

28,089 görüntüleme • 1 ay önce

Here is gemma-4-26B-A4B-it on A17 Pro chip w/8GB memory ( MacBook Neo) ~ 7 t/s running on AMX ( GPU is slower on A17) Gemma's 4 expert is x2.3 larger than Qwen See Qwen 35B below

Here is gemma-4-26B-A4B-it on A17 Pro chip w/8GB memory ( MacBook Neo) ~ 7 t/s running on AMX ( GPU is slower on A17) Gemma's 4 expert is x2.3 larger than Qwen See Qwen 35B below

Anemll

195,282 görüntüleme • 6 ay önce

The most amazing aspect of DeepSeek Open Source is that just the Reasoning Engine can be isolated and used with any other LLM. In fact you can have a mixture of Reasoning Engines cascade around a problem and then use an Operator type agent AI to function on the the results.

The most amazing aspect of DeepSeek Open Source is that just the Reasoning Engine can be isolated and used with any other LLM. In fact you can have a mixture of Reasoning Engines cascade around a problem and then use an Operator type agent AI to function on the the results.

Brian Roemmele

398,807 görüntüleme • 1 yıl önce

The M6 ANE is fast. We got Qwen3.8-27B running entirely on it with Vector Quantization (VQ) and up to 64K context, reaching ~18–40 tok/s with DFlash2, depending on context. VQ gives us ~2× on the ANE, and with M6 it’s finally practical for real models. This feels like the biggest step forward for the ANE in more than six years.

The M6 ANE is fast. We got Qwen3.8-27B running entirely on it with Vector Quantization (VQ) and up to 64K context, reaching ~18–40 tok/s with DFlash2, depending on context. VQ gives us ~2× on the ANE, and with M6 it’s finally practical for real models. This feels like the biggest step forward for the ANE in more than six years.

Anemll

37,066 görüntüleme • 13 gün önce

Apple is no longer hiding it. These machines are built for local AI. Apple unveiled the new Mac Mini with M6 their first 2nm chip. Starting at $899. What you can run on the $899 Mac Mini (32GB) 👇 → Qwen 3.6 27B : fits comfortably, near-GPT-4 quality → Qwen 3 Coder 32B : full coding assistant on your desk → Llama 4 8B at full precision → DeepSeek R1 32B for reasoning tasks → 13.5x faster LLM processing than M1. 4x faster than M4. What you can run on the Mac Studio with M5 Ultra (512GB) 👇 → Llama 4 70B at high quality quantization → DeepSeek V4 with hundreds of billions of parameters → Frontier-scale models entirely in local memory → Connect four Mac Studios over Thunderbolt 5 cluster them to run trillion-parameter models locally Apple said their Mac business grew 30% last quarter largely because people are buying these to run AI models on their own desks instead of paying for cloud tokens. The future of inference is local. Apple built the hardware for it.

Apple is no longer hiding it. These machines are built for local AI. Apple unveiled the new Mac Mini with M6 their first 2nm chip. Starting at $899. What you can run on the $899 Mac Mini (32GB) 👇 → Qwen 3.6 27B : fits comfortably, near-GPT-4 quality → Qwen 3 Coder 32B : full coding assistant on your desk → Llama 4 8B at full precision → DeepSeek R1 32B for reasoning tasks → 13.5x faster LLM processing than M1. 4x faster than M4. What you can run on the Mac Studio with M5 Ultra (512GB) 👇 → Llama 4 70B at high quality quantization → DeepSeek V4 with hundreds of billions of parameters → Frontier-scale models entirely in local memory → Connect four Mac Studios over Thunderbolt 5 cluster them to run trillion-parameter models locally Apple said their Mac business grew 30% last quarter largely because people are buying these to run AI models on their own desks instead of paying for cloud tokens. The future of inference is local. Apple built the hardware for it.

Vaibhav Sisinty

517,316 görüntüleme • 1 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.

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 • 7 ay önce

Reasoning VLAs can think. They just can't think fast. Until now. Introducing FlashDrive⚡ 🚀 716 ms → 159 ms on RTX PRO 6000 (up to 5.7×) ✅ Zero accuracy loss FlashDrive = streaming inference + DFlash speculative reasoning + ParoQuant W4A8 Real-time reasoning for autonomous driving is here!

Reasoning VLAs can think. They just can't think fast. Until now. Introducing FlashDrive⚡ 🚀 716 ms → 159 ms on RTX PRO 6000 (up to 5.7×) ✅ Zero accuracy loss FlashDrive = streaming inference + DFlash speculative reasoning + ParoQuant W4A8 Real-time reasoning for autonomous driving is here!

Zhijian Liu

191,948 görüntüleme • 5 ay önce

Alibaba’s Qwen 3.5 is now running fully on-device on the iPhone 17 Pro. It outperforms models 4x its size, delivers strong visual understanding, and lets you switch reasoning on or off. This demo uses the 2B 6-bit version, optimized with MLX for Apple Silicon.

Alibaba’s Qwen 3.5 is now running fully on-device on the iPhone 17 Pro. It outperforms models 4x its size, delivers strong visual understanding, and lets you switch reasoning on or off. This demo uses the 2B 6-bit version, optimized with MLX for Apple Silicon.

Hugging Models

203,208 görüntüleme • 7 ay önce

Mind = BLOWN 🤯 Grok 3 just launched with reasoning abilities that put other AIs to shame. Just spent hours testing it vs ChatGPT. My honest breakdown: • Reasoning mode destroys basic Grok for complex tasks • Image generation is SIGNIFICANTLY better than ChatGPT • Coding still needs work (ChatGPT wins here) • Includes web search similar to Perplexity ($16/mo for 50 searches) • New platform with sharable chats that index on Google Who's trying it?

Mind = BLOWN 🤯 Grok 3 just launched with reasoning abilities that put other AIs to shame. Just spent hours testing it vs ChatGPT. My honest breakdown: • Reasoning mode destroys basic Grok for complex tasks • Image generation is SIGNIFICANTLY better than ChatGPT • Coding still needs work (ChatGPT wins here) • Includes web search similar to Perplexity ($16/mo for 50 searches) • New platform with sharable chats that index on Google Who's trying it?

Julian Goldie SEO

11,517 görüntüleme • 1 yıl önce

3D-R1 is out on Hugging Face Enhancing Reasoning in 3D VLMs for Unified Scene Understanding 3D-R1 is an open-source generalist model that enhances the reasoning of 3D VLMs for unified scene understanding

3D-R1 is out on Hugging Face Enhancing Reasoning in 3D VLMs for Unified Scene Understanding 3D-R1 is an open-source generalist model that enhances the reasoning of 3D VLMs for unified scene understanding

AK

34,729 görüntüleme • 1 yıl önce

We've added a new "reasoning" focus (beta) on Perplexity for Pro users. It will use the new OpenAI o1-mini. There is no search integration yet. The model is slow, and usage is limited because of rate limits. It is good for puzzles, math, and coding. Pro searches are your best way to use multi-step reasoning + web search.

We've added a new "reasoning" focus (beta) on Perplexity for Pro users. It will use the new OpenAI o1-mini. There is no search integration yet. The model is slow, and usage is limited because of rate limits. It is good for puzzles, math, and coding. Pro searches are your best way to use multi-step reasoning + web search.

Aravind Srinivas

288,808 görüntüleme • 2 yıl önce

Wow Google really cooked with the Gemini live API! Audio/Video reasoning at the speed of light ⚡️ The second answer is actually off but I'm amazed at how quickly it can "think". It will only get better from here. Gradio demo on Hugging Face 🔜

Wow Google really cooked with the Gemini live API! Audio/Video reasoning at the speed of light ⚡️ The second answer is actually off but I'm amazed at how quickly it can "think". It will only get better from here. Gradio demo on Hugging Face 🔜

Freddy A Boulton

19,971 görüntüleme • 1 yıl önce

EBM are so back! Yann LeCun has been pointing here for years: AI reasoning needs systems that check structure before they answer. Aleph from Logical Intelligence now leads the major formal reasoning benchmarks – let me explain what it is -> 📺

EBM are so back! Yann LeCun has been pointing here for years: AI reasoning needs systems that check structure before they answer. Aleph from Logical Intelligence now leads the major formal reasoning benchmarks – let me explain what it is -> 📺

Ksenia Se

58,510 görüntüleme • 4 ay önce

Agents today are fragmented: keys on one chain, logic outside, wallet somewhere else. On Ritual, the agent lives on one substrate, can execute cross-chain, but its reasoning stays private. An agent that can derive its own keys and keep its own thoughts is actually autonomous.

Agents today are fragmented: keys on one chain, logic outside, wallet somewhere else. On Ritual, the agent lives on one substrate, can execute cross-chain, but its reasoning stays private. An agent that can derive its own keys and keep its own thoughts is actually autonomous.

Ritual Foundation

20,706 görüntüleme • 1 ay önce

i tested sonnet 5.5 on "code-to-painting". every pixel is generated by python the model wrote. it's shown a photo, writes a brush engine, renders, looks, and revises. you can see the large step-up from sonnet 5 in terms of coding + visual reasoning.

i tested sonnet 5.5 on "code-to-painting". every pixel is generated by python the model wrote. it's shown a photo, writes a brush engine, renders, looks, and revises. you can see the large step-up from sonnet 5 in terms of coding + visual reasoning.

Lance Martin

12,835 görüntüleme • 13 gün önce

You can now run Qwen 3.6 27B on <10GB of memory and retain 90% of the intelligence. Today, PrismML dropped Bonsai 27B - which takes their unique compression approach they have already successfully applied to Qwen3.6 8B and Flux2-Klein and applied it to a properly smart model. Here is it using an entire computer for deep research. Long horizon agentic calls already are strong with Qwen3.6 27B and they still are here - just on 10x less memory.

You can now run Qwen 3.6 27B on <10GB of memory and retain 90% of the intelligence. Today, PrismML dropped Bonsai 27B - which takes their unique compression approach they have already successfully applied to Qwen3.6 8B and Flux2-Klein and applied it to a properly smart model. Here is it using an entire computer for deep research. Long horizon agentic calls already are strong with Qwen3.6 27B and they still are here - just on 10x less memory.

Tim Carambat

63,449 görüntüleme • 2 ay önce