# Cogito

> Open-source LLM inference on AWS Trainium and NVIDIA Blackwell. 14× faster than typical providers.

- [Overview](https://docs.cogito.decart.ai/getting-started/overview.md): Stop watching your agent type. Cogito serves open-source LLMs 14× faster than typical providers, on AWS Trainium and NVIDIA Blackwell.
- [Quickstart](https://docs.cogito.decart.ai/getting-started/quickstart.md): Send your first Cogito request in under five minutes.
- [Authentication](https://docs.cogito.decart.ai/getting-started/authentication.md): API keys, header format, and best practices.
- [Model catalog](https://docs.cogito.decart.ai/getting-started/models.md): Available open-source models, with context windows, output caps, throughput, and pricing.
- [Pricing](https://docs.cogito.decart.ai/getting-started/pricing.md): How tokens are counted, billed, and discounted.
- [Streaming](https://docs.cogito.decart.ai/guides/streaming.md): SSE chunks, cancellation, error recovery.
- [Function calling](https://docs.cogito.decart.ai/guides/function-calling.md): Wire LLMs into your tools with reliable, parallel function calls.
- [Structured outputs](https://docs.cogito.decart.ai/guides/structured-outputs.md): Guaranteed JSON schema adherence on models that advertise structured outputs.
- [Auto-recharge](https://docs.cogito.decart.ai/billing/auto-recharge.md): Keep your balance topped up automatically so requests never 402 mid-deploy.
- [Kimi K2.6](https://docs.cogito.decart.ai/models/kimi-k2-6.md): Moonshot's flagship MoE for coding agents and long-horizon reasoning.
- [Kimi K2.7 Code](https://docs.cogito.decart.ai/models/kimi-k2-7.md): Moonshot's newest flagship MoE — the latest Kimi. Early-access / reserved capacity.
- [GLM-5.2](https://docs.cogito.decart.ai/models/glm-5-2.md): Zhipu's frontier open-weight MoE — 1M-token context and strong agentic reasoning.
- [GPT-OSS 120B](https://docs.cogito.decart.ai/models/gpt-oss-120b.md): OpenAI's open-weight 120B MoE. Strong general reasoning at near-commodity pricing.
- [Qwen3 235B](https://docs.cogito.decart.ai/models/qwen-3-235b.md): Alibaba's open-weight flagship — multilingual, tool-friendly, Apache 2.0.
- [Create chat completion](https://docs.cogito.decart.ai/api-reference/chat-completions.md): POST /v1/chat/completions
- [Create response](https://docs.cogito.decart.ai/api-reference/responses.md): POST /v1/responses
- [Create completion](https://docs.cogito.decart.ai/api-reference/completions.md): POST /v1/completions
- [List models](https://docs.cogito.decart.ai/api-reference/list-models.md): GET /v1/models
- [Errors](https://docs.cogito.decart.ai/api-reference/errors.md): Deterministic error schema. Always includes a request_id.

## Optional

- [Platform](https://cogito.decart.ai)
- [GitHub](https://github.com/DecartAI)