Model Hub

TinyLlama 1.1B chat — greedy decode contract. Published under its Hugging Face repository name (TinyLlama/TinyLlama-1.1B-Chat-v1.0); the earlier record TinyLlama/TinyLlama-1.1B-Chat carried an invented name and is hidden in favour of this one.

Qwen3 30B-A3B Thinking — Alibaba Qwen Mixture-of-Experts LLM (30B total, 3B activated per token) tuned for deep step-by-step reasoning with explicit thinking traces, excelling at math, code and logic. Apache 2.0.

DeepSeek-MoE 16B chat — fused-MoE runtime, greedy decode contract

IBM Granite 3.1 1B-A400M Instruct — a 32-expert mixture-of-experts language model (8 experts per token, 24 layers), the first MoE container through the fused expert dispatch; traced and validated on NVIDIA V100 in sequential, compiled and Triton modes