Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

RF-DETR by Roboflow now on MLX It can do realtime instance segmentation on-device and enable some cool use cases for visual analysis, monitoring and robotics like Reachy Mini. Also augmented VLM and VLA by preprocessing image and video with areas of interest. New release coming soon on mlx-vlm 🚀... show more

Prince Canuma

21,606 subscribers

30,499 views • 3 months ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Running Qwen2 VL 72B 4bit 🇨🇳 quantized model on Mac Mini M4 Pro 64GB. It works very well and delivers good accuracy, suitable for practical use cases. Integrating MLX and MLX-VLM backend into Sparrow as main solution for local inference and data extraction. Kudos to Prince Canuma for porting Qwen2 VL to MLX.

Running Qwen2 VL 72B 4bit 🇨🇳 quantized model on Mac Mini M4 Pro 64GB. It works very well and delivers good accuracy, suitable for practical use cases. Integrating MLX and MLX-VLM backend into Sparrow as main solution for local inference and data extraction. Kudos to Prince Canuma for porting Qwen2 VL to MLX.

Andrej Baranovskij

13,742 views • 1 year ago

oMLX is working really well as single machine inference engine for coding agents! Caching is managed perfectly (it can use a ton of disk space, be aware!) and oQ quantization delivers great results. Behind the scenes it uses the standard MLX building blocks (75% created by Prince Canuma 🙏): - mlx-lm - mlx-vlm - mlx-embeddings - mlx-audio I tested Qwen3.6-35B-A3B-oQ6 on M5 Max with two pi instances and it was fast and furious and leveraging cache like crazy as you can see in the video. Let me try to create some oQ versions (2,4,6?) of MiniMax M2.7 now and then I'll pass to distributed inference. I must win! 💪

oMLX is working really well as single machine inference engine for coding agents! Caching is managed perfectly (it can use a ton of disk space, be aware!) and oQ quantization delivers great results. Behind the scenes it uses the standard MLX building blocks (75% created by Prince Canuma 🙏): - mlx-lm - mlx-vlm - mlx-embeddings - mlx-audio I tested Qwen3.6-35B-A3B-oQ6 on M5 Max with two pi instances and it was fast and furious and leveraging cache like crazy as you can see in the video. Let me try to create some oQ versions (2,4,6?) of MiniMax M2.7 now and then I'll pass to distributed inference. I must win! 💪

Ivan Fioravanti ᯅ

19,384 views • 3 months ago

Next mlx-vlm release will ship with continuous batching support on the server 🚀 What's coming: → Continuous batching — new requests join the active batch immediately, no waiting. Mixed image + text batches supported → OpenAI-compatible API — field-for-field match with mlx-lm, reasoning/content split for thinking models, tag-aware streaming → Multi-turn tool calling — full tool use support across streaming and non-streaming, works with Gemma4 and other templates → Vision feature caching — cache image embeddings across turns. Gemma4: 228x speedup, Qwen3.5: 23x on cache hit All running locally on Apple Silicon. Check our this demo running 4 concurrent requests (mixed image + text) to gemma-4-26B-A4B-IT by Google Gemma in bf16 using Pi + MLX-VLM server on my M3 Ultra. One of the requests ingests a 8K resolution image!

Next mlx-vlm release will ship with continuous batching support on the server 🚀 What's coming: → Continuous batching — new requests join the active batch immediately, no waiting. Mixed image + text batches supported → OpenAI-compatible API — field-for-field match with mlx-lm, reasoning/content split for thinking models, tag-aware streaming → Multi-turn tool calling — full tool use support across streaming and non-streaming, works with Gemma4 and other templates → Vision feature caching — cache image embeddings across turns. Gemma4: 228x speedup, Qwen3.5: 23x on cache hit All running locally on Apple Silicon. Check our this demo running 4 concurrent requests (mixed image + text) to gemma-4-26B-A4B-IT by Google Gemma in bf16 using Pi + MLX-VLM server on my M3 Ultra. One of the requests ingests a 8K resolution image!

Prince Canuma

82,169 views • 3 months ago

Sam-Audio by AI at Meta is now on MLX-Audio 🔊 You can now isolate voices or sounds from any audio locally in your Mac. I added this simple UI to the MLX-Audio studio app to help you get started Note: - The release is coming soon for now please install from source. - Vision part was excluded for now. Let me know if you need it. - PR for UI:

Sam-Audio by AI at Meta is now on MLX-Audio 🔊 You can now isolate voices or sounds from any audio locally in your Mac. I added this simple UI to the MLX-Audio studio app to help you get started Note: - The release is coming soon for now please install from source. - Vision part was excluded for now. Let me know if you need it. - PR for UI:

Prince Canuma

19,027 views • 6 months ago

Congratulations to the Cohere team on releasing North Mini Code 1.0 🔥 We provide day-0 support on MLX thanks to our close collaboration with the Cohere team. North Mini Code 1.0 is a 30B param MoE model with 3B active. It runs at ~66 tok/s in BF16, truly impressive speeds before any compression. Here is a quick demo I built using Pi by Mario Zechner & Armin Ronacher ⇌ as my agent hardness. Get started today: 1. Install mlx-vlm from source: > uv pip install git+ 2. Start the server and power your favourite agents (pi, opencode, hermes and etc) > uv run mlx_vlm.server --model CohereLabs/North-Mini-Code-1.0 New release tomorrow

Congratulations to the Cohere team on releasing North Mini Code 1.0 🔥 We provide day-0 support on MLX thanks to our close collaboration with the Cohere team. North Mini Code 1.0 is a 30B param MoE model with 3B active. It runs at ~66 tok/s in BF16, truly impressive speeds before any compression. Here is a quick demo I built using Pi by Mario Zechner & Armin Ronacher ⇌ as my agent hardness. Get started today: 1. Install mlx-vlm from source: > uv pip install git+ 2. Start the server and power your favourite agents (pi, opencode, hermes and etc) > uv run mlx_vlm.server --model CohereLabs/North-Mini-Code-1.0 New release tomorrow

Prince Canuma

35,367 views • 1 month ago

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

3,555,682 views • 4 months ago

🗓️ Release Week Recap Big week. mlx-audio and mlx-vlm are now among some of the fastest-growing OSS projects. Here’s what we shipped last week. Gemma 4 on Apple Silicon Two awesome releases by our partner Google DeepMind & Google Gemma : > Gemma 4 12B — their new dense, unified multimodal model. It uses an encoder free audio path and simplified vision encoder. > Gemma 4 QAT — quantization-aware training checkpoints, optimized to run locally on consumer GPUs and edge devices, compressing the model while preserving the quality you expect from Gemma 4. On the audio 🎧 side we added support for 15+ new TTS, ASR & VAD models, faster long-form transcription, and an expanded OpenAI-compatible audio server. All local on Apple Silicon. Huge thanks to every contributor and my co-maintainer Lucas Newman. 🙏🏽

🗓️ Release Week Recap Big week. mlx-audio and mlx-vlm are now among some of the fastest-growing OSS projects. Here’s what we shipped last week. Gemma 4 on Apple Silicon Two awesome releases by our partner Google DeepMind & Google Gemma : > Gemma 4 12B — their new dense, unified multimodal model. It uses an encoder free audio path and simplified vision encoder. > Gemma 4 QAT — quantization-aware training checkpoints, optimized to run locally on consumer GPUs and edge devices, compressing the model while preserving the quality you expect from Gemma 4. On the audio 🎧 side we added support for 15+ new TTS, ASR & VAD models, faster long-form transcription, and an expanded OpenAI-compatible audio server. All local on Apple Silicon. Huge thanks to every contributor and my co-maintainer Lucas Newman. 🙏🏽

Prince Canuma

11,731 views • 1 month ago

Day 1 of 3 days of MLX: Introducing MLX-Audio-Swift SDK 🚀 A modular Swift SDK for voice agents and tasks on Apple Silicon built by Lucas Newman and yours truly. iOS, macOS, and visionOS developers can now build native apps with real-time, on-device audio intelligence: 🗣️ Text-to-Speech (TTS) 👂 Speech-to-Text (STT) 🔄 Speech-to-Speech (STS) 🎙️ Voice Activity Detection (VAD) and more. Only import the capabilities you need, nothing extra. Get started today and leave us a star ⭐️

Day 1 of 3 days of MLX: Introducing MLX-Audio-Swift SDK 🚀 A modular Swift SDK for voice agents and tasks on Apple Silicon built by Lucas Newman and yours truly. iOS, macOS, and visionOS developers can now build native apps with real-time, on-device audio intelligence: 🗣️ Text-to-Speech (TTS) 👂 Speech-to-Text (STT) 🔄 Speech-to-Speech (STS) 🎙️ Voice Activity Detection (VAD) and more. Only import the capabilities you need, nothing extra. Get started today and leave us a star ⭐️

Prince Canuma

160,927 views • 5 months ago

The latest MLX has a CUDA back-end! To get started: pip install "mlx[cuda]" With the same codebase you can develop locally, run your model on Apple silicon, or in the cloud on Nvidia GPUs. MLX is designed around Apple silicon - which has a unified memory architecture. It uses the same design with CUDA. So there's no need to move arrays around from CPU memory to GPU memory. Note, this is early days - some operations are missing and performance is still being optimized. But it's already quite fast for Transformer training, text generation, and more! Here's a demo using mlx-lm to generate text with Llama 3 8B (bf16) on an A100:

The latest MLX has a CUDA back-end! To get started: pip install "mlx[cuda]" With the same codebase you can develop locally, run your model on Apple silicon, or in the cloud on Nvidia GPUs. MLX is designed around Apple silicon - which has a unified memory architecture. It uses the same design with CUDA. So there's no need to move arrays around from CPU memory to GPU memory. Note, this is early days - some operations are missing and performance is still being optimized. But it's already quite fast for Transformer training, text generation, and more! Here's a demo using mlx-lm to generate text with Llama 3 8B (bf16) on an A100:

Awni Hannun

42,761 views • 1 year ago

SAM-2 is apache license - you can indeed just build crazy apps with them without fear. This weekend I ported SAM-2 models to run optimally on Apple Silicon. I have also quantized all the SAM models to 16bit, 8bit, and 4bit variants (collection below) Repo is open source and pip installable. Comes with all you'll need for interactive video segmentation - - bidirectional propagation, - positive and negative clicks, - bounding boxes, - streaming masks back to user, - lightweight fast-api server that you can extend I basically asked Codex to port over the original Facebook sam2 codebase to mlx and optimize it for mlx. And it just worked. This demo for example is a 4bit base-plus model (~100MB download) To speed up, you can also do spatial or temporal downsampling. That speeds inference on my old macbook by ~3x and the hit in parity is negligible.

SAM-2 is apache license - you can indeed just build crazy apps with them without fear. This weekend I ported SAM-2 models to run optimally on Apple Silicon. I have also quantized all the SAM models to 16bit, 8bit, and 4bit variants (collection below) Repo is open source and pip installable. Comes with all you'll need for interactive video segmentation - - bidirectional propagation, - positive and negative clicks, - bounding boxes, - streaming masks back to user, - lightweight fast-api server that you can extend I basically asked Codex to port over the original Facebook sam2 codebase to mlx and optimize it for mlx. And it just worked. This demo for example is a 4bit base-plus model (~100MB download) To speed up, you can also do spatial or temporal downsampling. That speeds inference on my old macbook by ~3x and the hit in parity is negligible.

AVB

17,363 views • 2 months ago

I made a few updates to the MLX port of F5 TTS over the holiday: - Longform generation is now supported, and text will be split automatically on sentence boundaries for natural-sounding speech. - You can now pipe in text from another process (e.g. a language model) and listen to the speech as each segment is generated. - I implemented RK4 sampling for the ODE, which achieves much better quality with fewer steps — it's now possible to generate speech at 1.6x realtime on my M3 Max. See the video of an example of generating speech in realtime from a language model entirely on-device! 🚀

I made a few updates to the MLX port of F5 TTS over the holiday: - Longform generation is now supported, and text will be split automatically on sentence boundaries for natural-sounding speech. - You can now pipe in text from another process (e.g. a language model) and listen to the speech as each segment is generated. - I implemented RK4 sampling for the ODE, which achieves much better quality with fewer steps — it's now possible to generate speech at 1.6x realtime on my M3 Max. See the video of an example of generating speech in realtime from a language model entirely on-device! 🚀

Lucas Newman

22,998 views • 1 year ago

If you let VLMs experiment on their own, they can do surprising things! From an image, we let a VLM code a 3D scene from scratch in Blender, and then render to verify/refine in a loop. Even when each step is imperfect, it gets results like this, with zero training.

If you let VLMs experiment on their own, they can do surprising things! From an image, we let a VLM code a 3D scene from scratch in Blender, and then render to verify/refine in a loop. Even when each step is imperfect, it gets results like this, with zero training.

Angjoo Kanazawa

51,928 views • 6 months ago

ParoQuant just got a big upgrade 🚀 ✅ Supports the new Qwen3.5 models ⚡ Now runs on MLX (fast local inference on Apple Silicon) 🧠 Preserves reasoning quality with 4-bit quantization We also built an agent demo running locally on my 4-year-old M2 Max. Can't wait to upgrade to an M5 Max and see what kind of magic we can do. ✨

ParoQuant just got a big upgrade 🚀 ✅ Supports the new Qwen3.5 models ⚡ Now runs on MLX (fast local inference on Apple Silicon) 🧠 Preserves reasoning quality with 4-bit quantization We also built an agent demo running locally on my 4-year-old M2 Max. Can't wait to upgrade to an M5 Max and see what kind of magic we can do. ✨

Zhijian Liu

49,358 views • 4 months ago

With the latest FilzaJailed 2.1, you can use it for TrollStore install on iOS 17.0 directly on device You need to use some application (eg: Esign) for unzip since Filza itself is broken on zip and unzip Here's the link: Here's the demo video:

With the latest FilzaJailed 2.1, you can use it for TrollStore install on iOS 17.0 directly on device You need to use some application (eg: Esign) for unzip since Filza itself is broken on zip and unzip Here's the link: Here's the demo video:

Huy Nguyen

54,497 views • 3 months ago

LiteParse Samples 📄🧑‍🏫 LiteParse is our free and fast document parser that can parse text with bounding boxes from any document. I’ve created a new repository that contains some demos on how you can make use of its outputs! ✅ Comparison against PyPDF and PyMuPDF ✅ Visual Citations: search for any keyword and see joined bounding boxes pop up in the source images It should be a default tool to help AI agents OCR any document type (supports PDF, Word, Powerpoint + dozens of other formats), before heavier-weight VLM-based document parsing solutions (like our own LlamaParse). LiteParse Samples repo: LiteParse:

LiteParse Samples 📄🧑‍🏫 LiteParse is our free and fast document parser that can parse text with bounding boxes from any document. I’ve created a new repository that contains some demos on how you can make use of its outputs! ✅ Comparison against PyPDF and PyMuPDF ✅ Visual Citations: search for any keyword and see joined bounding boxes pop up in the source images It should be a default tool to help AI agents OCR any document type (supports PDF, Word, Powerpoint + dozens of other formats), before heavier-weight VLM-based document parsing solutions (like our own LlamaParse). LiteParse Samples repo: LiteParse:

Jerry Liu

13,196 views • 3 months ago

You can now vibecode your own WisprFlow or Monologue alternative that runs completely locally on Apple Silicon using MLX-Audio-Swift 🔥 Check out this live transcription of Dwarkesh Patel interview with Andrej Karpathy using Qwen3-ASR-0.6B quantized to 4bit on a M3 Max. It also runs in realtime on a iPhone 15 Pro and iPad Pro M1. No cloud. No API keys.

You can now vibecode your own WisprFlow or Monologue alternative that runs completely locally on Apple Silicon using MLX-Audio-Swift 🔥 Check out this live transcription of Dwarkesh Patel interview with Andrej Karpathy using Qwen3-ASR-0.6B quantized to 4bit on a M3 Max. It also runs in realtime on a iPhone 15 Pro and iPad Pro M1. No cloud. No API keys.

Prince Canuma

60,974 views • 5 months ago

DeepSeek-v4 now runs at ~23-26 tok/s on MLX! I made some custom kernels for the sinkhorn and it took gen speeds for 17 -> 26 tok/s. The weights are also significantly smaller thanks to Pedro Cuenca tip about keeping the experts in MXFP4! Now you can use it to power your local coding agents (PI, Open code, Hermes agent or even CC) PR:

DeepSeek-v4 now runs at ~23-26 tok/s on MLX! I made some custom kernels for the sinkhorn and it took gen speeds for 17 -> 26 tok/s. The weights are also significantly smaller thanks to Pedro Cuenca tip about keeping the experts in MXFP4! Now you can use it to power your local coding agents (PI, Open code, Hermes agent or even CC) PR:

Prince Canuma

58,451 views • 3 months ago

📢 OneCanvas: 3D Scene Understanding via Panoramic Reprojection We extract features from video frames and reproject them into one occlusion-free view of the whole scene that a 2D VLM reads just like a normal image. We can center this view on any viewpoint, including an agent's own pose for situated reasoning. The same projection lets us create spatial training tasks with no human annotation, solvable only by reasoning over the 3D positions of real object features placed on an otherwise empty canvas. The result is a stock 2D VLM that reasons in 3D, setting a new state of the art across spatial benchmarks at far less compute. 🌐 ▶️ Great work by Bartłomiej Baranowski & Dave Zhenyu Chen

📢 OneCanvas: 3D Scene Understanding via Panoramic Reprojection We extract features from video frames and reproject them into one occlusion-free view of the whole scene that a 2D VLM reads just like a normal image. We can center this view on any viewpoint, including an agent's own pose for situated reasoning. The same projection lets us create spatial training tasks with no human annotation, solvable only by reasoning over the 3D positions of real object features placed on an otherwise empty canvas. The result is a stock 2D VLM that reasons in 3D, setting a new state of the art across spatial benchmarks at far less compute. 🌐 ▶️ Great work by Bartłomiej Baranowski & Dave Zhenyu Chen

Matthias Niessner

25,120 views • 1 month ago

QVAC SDK 0.12.0 is now live, bringing longer context, increased memory optimisation, new modalities, and broader ecosystem support directly to your device. Key Features and Updates: - TurboQuant KV-Cache Quantization: Fit much longer context in the same memory. TurboQuant, an algorithm from Google Research, compresses the KV cache by up to 5x, near-lossless. - Text-to-Video: Generate video from a text prompt, fully local, with the new wan2.1 model in the Diffusion addon - Apple Metal Performance for Flux2-klein: Diffusion on Apple Silicon now matches MLX performance, the native benchmark for Apple GPUs - Robot Control (new VLA addon): A GGML-based Vision-Language-Action addon brings fast, efficient robot control to edge devices - Coding Assistant / Harness Support: QVAC now works with OpenCode and OpenClaw as a local provider. A new @qvac/ai-sdk-provider package automates model registry and provider integration - Cross-Platform Voice: Text-to-speech and Parakeet transcription moved from ONNX to the GGML engine for better CPU and GPU support on macOS, iOS, Windows, Linux, and Android. Parakeet also adds long-term streaming diarization (tracking who spoke when on live audio) - Faster Lightweight Visual Classification: A new GGML-based Classification addon delivers millisecond-level classification, useful where a vision-language model (VLM) would be unnecessarily slow - Under the Hood: Fabric synced to llama.cpp v8828 (from v8189), plus GPU acceleration added to image-upscale models for faster results Full release notes:

QVAC SDK 0.12.0 is now live, bringing longer context, increased memory optimisation, new modalities, and broader ecosystem support directly to your device. Key Features and Updates: - TurboQuant KV-Cache Quantization: Fit much longer context in the same memory. TurboQuant, an algorithm from Google Research, compresses the KV cache by up to 5x, near-lossless. - Text-to-Video: Generate video from a text prompt, fully local, with the new wan2.1 model in the Diffusion addon - Apple Metal Performance for Flux2-klein: Diffusion on Apple Silicon now matches MLX performance, the native benchmark for Apple GPUs - Robot Control (new VLA addon): A GGML-based Vision-Language-Action addon brings fast, efficient robot control to edge devices - Coding Assistant / Harness Support: QVAC now works with OpenCode and OpenClaw as a local provider. A new @qvac/ai-sdk-provider package automates model registry and provider integration - Cross-Platform Voice: Text-to-speech and Parakeet transcription moved from ONNX to the GGML engine for better CPU and GPU support on macOS, iOS, Windows, Linux, and Android. Parakeet also adds long-term streaming diarization (tracking who spoke when on live audio) - Faster Lightweight Visual Classification: A new GGML-based Classification addon delivers millisecond-level classification, useful where a vision-language model (VLM) would be unnecessarily slow - Under the Hood: Fabric synced to llama.cpp v8828 (from v8189), plus GPU acceleration added to image-upscale models for faster results Full release notes:

QVAC

9,932,369 views • 1 month ago