System 1 decision models (Jev, Laya, Cua-S1) as brain for agents: Browser use, computer use, games and robotics

community maintained vllm/vllm-omni style inference engine for system1 models in Rust for extreme efficient performance

A framework for efficient model inference with omni-modality models

A SOTA quantization toolkit for high-accuracy low-bit LLM inference|简洁且高效的量化工具包

Qwen3.8-27B on a single RTX 3090 with vLLM: ~1,000 tok/s at 64 concurrent (int8 tensor-core GEMMs, fp16 DeltaNet state), ~114 tok/s single-user at default sampling / ~124 greedy (MTP drafts, own-ou…

High-performance RL post-training infrastructure. Designed to achieve bitwise operator-level train-inference consistency across heterogeneous engines and extreme memory efficiency for GRPO, PPO, etc.

Train transformer language models with reinforcement learning.

🧠「大模型」2小时完全从0训练64M的小参数LLM！Train a 64M-parameter LLM from scratch in just 2h!