Inception released Mercury 3, its most capable model to date. It is a diffusion-based language model suited for agentic workflows and real-time applications. Its distinguishing trait is speed, over 40x larger than Mercury 1 while generating at the same throughput. It produces and refines tokens in parallel rather than one at a time. It supports a 1,000,000 token context window. It costs $0.10 per 1M input tokens and $0.35 per 1M output tokens. It supports tool calling and structured outputs. It is hosted by OpenRouter and not available through EU in-region routing. The model was released on October 16, 2026.
What changed
Mercury 2 was the lab's most capable model.
What it unlocks
Running agentic workflows and real-time applications with lower latency.
- 40x larger than Mercury 1
- 1M token context
- $0.10 per 1M input tokens
- $0.35 per 1M output tokens
What you need to act on it
- access via OpenRouter
- not available through EU in-region routing
Sources