Awni Hannun / @awnihannun · Scour

When you have a question that you are about to ask AI

twitter.macworks.dev

·2d

RT by @awnihannun: Microsoft's MIT licensed VibeVoice speech-to-text model (think Whisper with speaker diarization) is really good - my notes on running the 5.7...

twitter.macworks.dev

·1w

RT by @awnihannun: Finally... then mlc benchmarking leaderboard is online:

twitter.macworks.dev

·1w

RT by @awnihannun: After a small patch to mlx-lm, I have tool parameter streaming \o/

twitter.macworks.dev

·1w

Adopting Claude speak in my regular life, episode 1:

twitter.macworks.dev

·1w

RT by @awnihannun: 🚩First time to release MLX model quantized by AutoRound:

twitter.macworks.dev

·1w

RT by @awnihannun: In the new release of MLX we are bringing thread safety:

twitter.macworks.dev

·1w

RT by @awnihannun: MLX's implementation of RDMA (Remote Direct Memory Access) over Thunderbolt on macOS, can now be used as an independent library by anyone:

twitter.macworks.dev

·1w

The biggest companies can’t be just hardware or just software or just AI. They have to be extreme co-design companies (like Nvidia, but even more). That’s where...

twitter.macworks.dev

·2w

RT by @awnihannun: Kimi K2.6 was released 1h ago, and it looks amazing!

twitter.macworks.dev

·2w

RT by @awnihannun: Introducing the MLX-Benchmark Suite!!

twitter.macworks.dev

·2w

RT by @awnihannun: Qwen3.6 4bit DWQ now up on MLX, uses custom quantization scheme (4bit MLP 8bit everything else) + DWQ for additional gains. It gets 0.0225 KL...

twitter.macworks.dev

·2w

Link: https://huggingface.co/blog/transformers-to-mlx

twitter.macworks.dev

·2w

Great blogpost from @pcuenq on making a new skill + test harness to automate porting new models from Transformers to mlx-lm

twitter.macworks.dev

·2w

RT by @awnihannun: 🔈 Every model added to transformers has to be available on Apple Silicon 🍎 at once. We built a Skill and test harness for mlx-lm to get us cl...

twitter.macworks.dev

·2w

RT by @awnihannun: Google's Gemma 4 26B A4B running on MacBook Pro M5 Max

twitter.macworks.dev·2w

RT by @awnihannun: I open-sourced DDTree-MLX: tree-based speculative decoding for Apple Silicon.

twitter.macworks.dev

·2w

RT by @awnihannun: 🔥 DFlash x MLX is happening!

twitter.macworks.dev

·2w

RT by @awnihannun: dflash-mlx v0.1.1

twitter.macworks.dev

·3w

RT by @awnihannun: Medical AI models now run on iPhone. No cloud. No API.

twitter.macworks.dev

·3w

Log in to enable infinite scrolling