AI Models

Anthropic's newest frontier flagship — the strongest Claude ever shipped, for the problems nothing else can solve. 1M-token context.

Anthropic's balanced flagship — near-Opus intelligence at a fraction of the cost. Chat with it right here in your browser, or wire it into your app through one OpenAI-compatible endpoint.

Anthropic's most powerful production model — state-of-the-art coding, agentic workflows and hard reasoning, with a 1M-token context window.

The proven previous-generation Opus flagship — Opus-class reasoning and coding with adaptive thinking and a 1M-token context window.

Opus-class intelligence with visible extended thinking — watch the model reason step by step before it answers. 1M-token context.

The balanced previous-generation Sonnet — strong everyday intelligence with visible extended thinking and a 1M-token context window.

The fastest and most affordable Claude — near-instant responses for high-volume, latency-sensitive workloads, with 200K context.

The flagship tier of OpenAI's GPT-5.6 family — maximum reasoning power for the hardest coding, analysis and agent tasks, with 1M context.

The balanced mid-tier of OpenAI's GPT-5.6 family — strong general reasoning and coding with 1M context at a fair price.

OpenAI's light GPT-5.6 tier — everyday intelligence with a full 1M-token context window at a remarkably small price.

OpenAI's flagship text-to-image model — turn a sentence into a polished, photorealistic image with reliable text rendering. Up to 4K.

Prompt-driven image editing from OpenAI — restyle, replace backgrounds, add or remove objects while keeping the subject intact.

The newest generation of Google's efficient line — stronger reasoning than 3.5 Flash at the same class of speed and cost. 1M context.

Google's Pro-tier model — deep reasoning and strong coding beyond the Flash line, with native multimodality and 1M context.

Google's high-efficiency multimodal model — near-Pro reasoning at Flash-class speed and price, with tunable thinking levels and 1M context.

The lightest tier of Google's Gemini line — built for massive-scale pipelines where unit cost rules, still with a 1M-token context window.

Moonshot AI's flagship open-weight multimodal reasoning model — built for complex coding, long-horizon agents and image-grounded tasks, with 1M context.

Moonshot AI's proven coding and agent workhorse — strong tool use and visible reasoning at a very competitive price, with 256K context.

DeepSeek's 1.6T-parameter MoE open-source flagship — frontier-adjacent reasoning and coding at a fraction of flagship prices, with 1M context.

DeepSeek's efficiency model — remarkably capable output at rock-bottom per-token cost, with automatic caching and 1M context.

Zhipu AI's flagship model — strong coding, tool use and multilingual performance with a 1M-token context window at a value price.

xAI's newest flagship model — an upgrade over Grok 4.5 in reasoning and coding, with the same famously direct voice. 500K context, automatic caching.

xAI's flagship model — sharp reasoning and strong coding with a famously direct voice. 500K context and automatic caching.

Alibaba's flagship Qwen model — top-tier reasoning, coding and standout multilingual ability with a 1M-token context window.

The balanced mid-tier of Alibaba's Qwen line — strong reasoning and multilingual quality at a value price, with a 1M-token context window.

Alibaba's coding-specialized Qwen — tuned for software engineering, tool use and agent workflows, with a 256K-token context window.

Google's top-tier image model — studio-grade fidelity, advanced text rendering and multi-image composition, built on Gemini 3 Pro image tech.

Google's Nano Banana 2 — fast, budget-friendly image generation with strong text rendering and up to 4K, built on Gemini image tech.

MiniMax's flagship multimodal video model: native 2K clips up to 15 seconds from text, images and reference media — at prices well below the official rate card.

ByteDance's Seedance 2.0 video model — clips up to 15s from text, images and references.

ByteDance's Seedance 2.0 video model — clips up to 15s from text, images and references.

ByteDance's Seedance 2.0 video model — clips up to 15s from text, images and references.