Inception Labs has released Mercury 3, the fourth model in its series, which utilizes a diffusion-based architecture to generate tokens in parallel. The model achieves speeds of 1,068 tokens per second on NVIDIA GPUs and features a 1-million-token context window.
The model supports tool calling, structured outputs, and tunable reasoning. According to the company, it matches or outperforms GPT-6 Luna and Claude Haiku 5.5 on 8 to 9 out of 12 benchmarks, while operating 4 to 8 times faster than these competitors.
Mercury 3 is available via API on the Venice.ai platform with compatibility for OpenAI systems. Developers can access it using the model ID mercury-3. The company is targeting use cases involving low-latency requirements, such as AI agents and real-time search.
List pricing is set at $0.20 per million input tokens and $0.70 per million output tokens. A temporary 50% launch discount reduces these rates to $0.10 and $0.35 per million tokens respectively.
