Skip to content

Fields

Open source Latest

Open-weight models you can download and self-host, open-source frameworks and AI projects trending on GitHub.

62 picked45 in the last 30 days433 collected

November 19 01:27 updated

Open source · Picks · page 4

February 11WednesdayItems 61–62 of 62
  1. Modal Blog78

    Modal launches free GLM-5 endpoint and publishes self-deployment and frontend integration guides

    Modal has launched a free OpenAI API-compatible endpoint for GLM-5, available through the end of April, limited to a single concurrent request, and has also released reproducible code for deploying it with SGLang on a single node with 8 B200 GPUs.

    Why it matters: Modal provides a free OpenAI-compatible endpoint for Z.ai's GLM-5 and gives specific configurations for integrating it in OpenCode, OpenClaw, Claude Code, and the Vercel AI SDK.

September 26, 2025Friday
  1. Modal Blog60

    Modal engineers reverse-engineer how Flash Attention 4 works

    Modal engineers read the released Flash Attention 4 source code and reverse-engineered how the Blackwell-optimized attention kernel works, since no official technical report exists yet. The write-up traces a tile through the kernel's warp-specialized asynchronous pipeline and details two math tricks: a cubic-polynomial software exponential that can replace SFU-based exponentiation, and a smarter online softmax rescaling rule that Tri Dao said cuts corrections by a factor of 10. FA4 is reported to be roughly 20% faster than Nvidia's closed-source cudnn attention kernels, and a backwards version was not yet available at the time of writing.

    Why it matters: Modal engineers read the released Flash Attention 4 source and explain its warp-specialized pipeline and two math tricks, offering a transferable view of where CUDA kernel design is heading.