AI Models

Picks a capable model for you and keeps going if a provider is overloaded.

Our most capable model — frontier-level reasoning for complex agentic work, long-horizon tasks, and enterprise projects

The best combination of speed and intelligence, with adaptive thinking and long context support

Fast and intelligent model for quick tasks

Google's most advanced Gemini 3.1 model with powerful agentic capabilities, multimodal understanding, and state-of-the-art reasoning.

Google's workhorse Flash model — stronger agentic and multimodal performance than 3.5 Flash with reduced token usage on long-horizon coding and reasoning tasks, at a lower output price. Dynamic thinking is on by default.

Mistral's most capable model — strong complex reasoning and multimodal understanding. Ahead of Mistral Large 3 despite the smaller-sounding name.

Open-weight model with strong multilingual support, at a third the price of Mistral Medium 3.5 — a step below it on complex reasoning.

Hybrid model optimized for general chat, coding, agentic tasks, and complex reasoning with text and image input support

OpenAI's new flagship — token-efficient with stronger frontend design judgment, programmatic tool calling, and max reasoning effort for the hardest tasks. Priced like Claude Opus; available on the Pro plan and above.

GPT-5.6 Terra delivers strong performance at a lower price than Sol, with a 1M context window and optional reasoning. Sonnet-equivalent OpenAI option for Basic+ users.

GPT 5.6 Luna is an efficient, high-volume version of GPT-5.6 — fast and cost-effective for everyday tasks.

OpenAI's open-weights 120B-parameter model.

xAI's newest flagship, taking the top spot on independent agentic tool-use benchmarks — though its 500K token context window trades size for reasoning strength versus Grok 4.3's 1M.

xAI's fast model with an unprecedented 2 million token context window.

DeepSeek's flagship V4 reasoning model with strong coding and math capabilities and a 1 million token context window.

DeepSeek's fast and efficient V4 model optimized for speed with a 1 million token context window.

Moonshot AI's new flagship — a 2.8T-parameter model with native vision input and long-horizon agentic reasoning across a 1 million token context window.

A lower-latency, higher-throughput serving tier of GLM-5.2 for faster responses at a higher price.

Z-AI's general-purpose model with strong multilingual capabilities and a 1 million token context window.