Stars
Monitoring platform for ML teams building real‑world AI at scale.
Auto researching harness that burns through everything.
Drop-in OpenAI-compatible cache that replays repeated computer-use and agent workflows at near-zero cost.
Connect your agents & product to 1,000 APIs.
A native macOS auto-research harness for Codex and Claude Code. Pick an open research project and press Play.
Deployment & experimentation of LLM inference at production scale (P/D disaggregation, KV offloading, parallelism, and autoscaling).
AI coding agent that makes a team of small models beat a frontier one
🔥 Open Source Browser API for AI Agents & Apps. Steel Browser is a batteries-included browser sandbox that lets you automate the web without worrying about infrastructure.
Browser automation CLI for AI agents
Batched DeepSeek-V4.1-Flash inference on 8x V100 (sm_70): low-bit lookup formats, split experts, NVLink collectives
LLM speculative inference server for heterogeneous hardware & consumer GPUs
Run GLM-5.3-Flash (321B), DeepSeek-V4.1-Flash (748B) and MiniMax-M3.1-Flash (428B) on your own GPUs - one engine, one dashboard, OpenAI- and Anthropic-compatible API.
Xolotl: run DeepSeek-V4.1-Flash, all 552B, as a local coding agent on one gaming PC (RTX 4090 + 110 GB RAM): ~33 tok/s on agent tasks, ~41 in fast mode, 512K context. Plus GLM-5.3-Flash, Qwen3.8 an…
Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.
The open agent skills tool - npx skills
A hierarchical inference engine for frontier MoE models: hot experts on the GPU (CUDA, Vulkan), warm ones in RAM, cold ones streamed from NVMe. Written in C, runs on the hardware you already own.
DeepSeek-V4.1-Flash-1017 (552B MoE + 196B Engram) on 2x RTX 3090 + 128 GB RAM. Rust + CUDA, T16 quantization. 40+ tok/s measured with expert deferral + FP8 projections.
SolidCraft: an open-source parametric 3D CAD application built in pure Rust. Desktop, web, and automation through MCP.
Feather agent harness, training integrations, and evaluation tools for knowledge work
Unified Schema-Based Information Extraction
HVM1 (2022): a massively parallel, optimal functional runtime in Rust
Qwen3.5 inference server on hand-written CUDA kernels with continuous batching
Serve large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127 tok/s single-user (381 when the answer quotes the prompt), ~1,035 tok/s at 64 concurrent…
The AI Compute Platform for frontier teams. SkyPilot turns fragmented AI compute into one AI supercomputer, so frontier AI teams build custom intelligence faster.
Easily fine-tune (SFT/RL), evaluate, and deploy Qwen, Gemma, or any open agentic LLM/VLM!
Make power hour drinking-game videos: one minute per song, drink signals, overlays, trivia mode. Windows desktop app (Flask + pywebview + FFmpeg).