SignalTech/AI
Inception "Mercury 3", 1000 tok/s diffusion LLM
High-priority alert based on source significance and keyword relevance.
Context
What happened: Inception released Mercury 3, a diffusion-based LLM it says outperforms GPT-6 Luna and Claude Haiku 5.5 on benchmarks while running 4 to 8x faster. It is available on the Inception API Platform and through OpenRouter, Baseten, and Vercel.
Key numbers: 1,068 tokens per second on NVIDIA GPUs; 1M token context, up from
Why it matters: Inception says the scaling laws that drove autoregressive progress hold for diffusion, with intelligence up 3x from Mercury 1 (mini) to Mercury 3. Early-access partners cite latency gains in browser agents, tutoring, search, and customer support.
What changed
Inception released Mercury 3, a diffusion-based LLM it says outperforms GPT-6 Luna and Claude Haiku 5.5 on benchmarks while running 4 to 8x faster. It is available on the Inception API Platform and through OpenRouter, Baseten, and Vercel.
Why it matters
Inception says the scaling laws that drove autoregressive progress hold for diffusion, with intelligence up 3x from Mercury 1 (mini) to Mercury 3. Early-access partners cite latency gains in browser agents, tutoring, search, and customer support.