Open source and routing billions of tokens across the AI ecosystem
Meet ARC
A model that chooses the model.
ARC selects the model and reasoning effort most likely to succeed at every turn—at the lowest total cost.
- Turn-level
- Cache-aware
- Open weights soon
- DeepSeek V4 Flash Non-thinking
- DeepSeek V4 Flash Thinking
- GPT-5.6 Sol xhigh
- Qwen3.6-35B-A3B Local
Build with adaptive intelligence.
Use ARC directly, or through products designed around it.
The model that chooses the model.
ARC works behind the API you already use, matching every request to the intelligence it needs.
- AnthropicMessages
- OpenAIResponses
- yourharness
ARC
adaptive allocation
- Claude
- OpenAI
- DeepSeek
- GLM
- Kimi
- Qwen
- Gemini
- Gemma
- Mistral
Build four apps in parallel. Pick the best.
Four working versions of the same idea—compare them, then keep building from the one you choose.
Build with ARC in 
Production-ready software from an agent that adapts to the work.
- Cloud in the browser, Desktop on your machine
- Writes, runs, and iterates on real code
- ARC picks the model on every turn
Workshop Desktop
