
Inception's Mercury 2.5: Diffusion LLM Hits 1,100 Tokens/sec, Claims Frontier-Level Quality
Inception Labs releases Mercury 2.5, a diffusion-based LLM boasting 40% quality gains over Mercury 2, 1,107 tokens/sec throughput, and pricing that undercuts mainstream models. The update targets latency-critical workloads like voice agents and coding subagents.









