This saves all GPU poor 😃 #llms #mlx #ChatGPT #Startup

Green-Bit-LLM github.com/GreenBitAI/gre…

Happy to share DeepMixtral-8x7b-Instruct. A direct extraction/transfer of Mixtral Instruct's experts into Deepseek's architecture. Performance is identical, if not even a bit better, and seems more malleable to training. Collaborators Eric Hartford Fernando Fernandes Neto.

Got some nice pictures from GPT-4 👉huggingface.co/GreenBitAI 👉github.com/GreenBitAI/gre…

Honestly my thought currently running 2 production AI SaaS apps! 1. Anthropic's Claude is too weak - Ask it to be sarcastic it say's I can't hurt sentiments etc. Bruh! Lacking JSON mode is a bummer. You can try workarounds but any JSON with nesting is a pain! 2. Even if open

pip install -U mlx

It is good to know that apple's edge deployment uutilizes the same low-bit layout (a mix of INT4/INT2 ) as green-bit-llm. huggingface.co/blog/NicoNico/…

I'm super excited about M5. It's going to help a lot with compute-bound workloads in MLX. For example: - Much faster prefill. In other words time-to-first-token will go down. - Faster image / video generation - Faster fine-tuning (LoRA or otherwise) - Higher throughput for

LoRA fine-tuning Qwen3 4B on the DGX spark with mlx / mlx-lm. Gets a very respectable ~1200 tok/sec.

Building AgentCraft v1 with AgentCraft v0 is 🤌 Managed up to 9 Claude Code agents with the RTS interface so far. There's a lot to explore, but it feels right. v1 coming soon