INCEPTION SAYS ITS MERCURY 3 MODEL BEATS GPT - 6 LUNA AND RUNS FAR FASTER

What happened: Inception released Mercury 3, a diffusion LLM that refines many tokens in parallel rather than one at a time. The company says it beats OpenAI's GPT-6 Luna, Anthropic's Claude Haiku 5.5 and Zhipu's GLM-5.3 Flash on its own benchmarks. Key numbers: Self-reported speed of more than 1,000 tokens a second, four to eight times faster than rivals. Why it matters: Inception pitches speed over price for agents and coding tools that call a model repeatedly, and the launch pricing undercuts rivals on output tokens. No outside group has published a test of the model yet.

Inception released Mercury 3, a diffusion LLM that refines many tokens in parallel rather than one at a time. The company says it beats OpenAI's GPT-6 Luna, Anthropic's Claude Haiku 5.5 and Zhipu's GLM-5.3 Flash on its own benchmarks.

Inception pitches speed over price for agents and coding tools that call a model repeatedly, and the launch pricing undercuts rivals on output tokens. No outside group has published a test of the model yet.