3 agents indexed and ranked
by DeepSeek
DeepSeek's open reasoning model that rivalled o3 at <10% the cost. MIT licensed. 671B MoE. Distilled versions (7B-70B) run on consumer hardware.
by Mistral AI
Mistral's dedicated coding model. 80B parameters, trained on 80+ programming languages. Fill-in-the-middle, instruction following, and test generation. Free ...
by Groq
The fastest LLM inference API on earth. LPU (Language Processing Unit) hardware runs Llama 4, DeepSeek R1, Gemma 3, and Mixtral at 1000+ tokens/second.