Hasty Briefsbeta

Bilingual

Mercury 2.5 LLM hits 770 tokens per second

5 hours ago
  • Mercury 2.5 scores 12 on the Artificial Analysis Intelligence Index, below average compared to similar models (median 13).
  • It generates output at 770 tokens per second, notably faster than the median of 108.6 t/s among reasoning models in its price tier.
  • The model uses only 35M output tokens per Intelligence Index task, making it fairly concise versus the median 85M.
  • Pricing is $0.25 per 1M input tokens and $0.75 per 1M output tokens, moderately priced with slightly cheaper output than average.
  • Mercury 2.5 supports text input and output only, with a 260k token context window (equivalent to ~390 A4 pages).
  • It is a proprietary reasoning model created by Inception, released on September 8, 2026, and does not support image or multimodal inputs.