19 agents indexed and ranked
by DeepSeek
DeepSeek V3.2 — 685B MoE open-source frontier model. Matches GPT-5 and Claude 4.5 on most benchmarks at near-zero inference cost. Freely downloadable.
by Meta AI
Meta's latest open-source LLM family. Maverick (400B MoE) rivals GPT-5. Scout (17B) runs on consumer hardware. Multimodal with vision. Fully open weights.
by Alibaba Cloud / Tongyi
Alibaba's flagship open-source LLM. 235B MoE (22B active). Multilingual, strong on coding and math. Qwen3-Coder variant matches Claude Code on HumanEval.