Mercury 3 now available on AI Gateway
By Steven Van ·
Inception's diffusion model offers 1M-token context and over 1,000 tokens per second, at 50% off launch pricing on AI Gateway.
Mercury 3, a diffusion-based language model from Inception, is now available on Vercel AI Gateway, with 50% off launch pricing. The model name is inception/mercury-3.
What the model does
Mercury 3 uses diffusion to generate text through parallel refinement. Inception reports strong results across a range of benchmarks, with generation speeds above 1,000 tokens per second. Vercel says it brings improved reasoning to latency-sensitive workloads such as browser agents, real-time search, interactive education and customer support. A 1M-token context window is meant to give those applications room for long documents and extended conversation histories.
Using it
Set AI_GATEWAY_API_KEY in your environment and pass inception/mercury-3 as the model name. The release note's example uses the AI SDK's streamText with that model ID. Examples for Chat Completions and fx are also shown there. The model can also be tried in the model playground.
In coding agents
To use it in Claude Code, Codex, Cursor and other agents, install the latest Vercel CLI and run npx vercel ai-gateway setup. Then select inception/mercury-3 in the agent. The coding agents guide covers the setup. The command detects installed agents and shows a diff of each planned change before writing anything.
Pricing
AI Gateway reflects provider pricing with no markup. It does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests.