10 models.
One stable API.
OpenRouter has 400+. We picked the ones worth using and verify them weekly. No dead aliases, no surprise deprecations, no "model not found."
DeepSeek V3
DefaultBest valuedeepseek-v3The default. 90% cheaper than GPT-4o, near-identical capability on most tasks.
Gemini 2.5 Flash
1M contextMultimodalgemini-flashMillion-token context. Best for long documents, code analysis, RAG.
Claude Opus 4.8
FrontierReasoningclaude-opus-4-8Anthropic frontier. For when capability matters more than cost.
Grok 4.20
2M contextNewgrok-4-202-million-token context. The largest context window we offer.
Kimi K2-thinking
ThinkingAgentickimi-k2-thinkingMoE with extended thinking. Strong on agentic + tool-use loops.
MiniMax M2.5
Cost-effectiveminimax-m2.5Strong cost/quality ratio for general-purpose chat + summarization.
GLM 4.7
CodingTool useglm-4.7Best-in-class for structured outputs and tool calling.
Llama 3.3 70B
FreeOpen weightsllama-3.3-70b-instructFree tier. Strong baseline 70B model. Great for prototyping.
Qwen3 Coder
FreeCodingqwen3-coderFree coding-specialist. Tuned for code generation + completion.
Nemotron Super
FreeNVIDIAnemotron-superNVIDIA Nemotron Super. Free tier, strong general capability.
Want a model we don't have?
Tell us. If it's available through one of our underlying providers we'll add it.
Request a model