LaunchAI ModelsSeptember 29, 2026

Anthropic's Opus 5.5 leads SimpleBench at 88.4%, ships alongside GPT-6 Sol and Luna

Read original source →latent.space

Opus 5.5 leads SimpleBench at 88.4% and ranks as Anthropic's best vision model yet, at roughly 60% lower cost than Fable 5.1. On Terminal-Bench-Science it climbs from 24% at low reasoning effort to 62% at xhigh, then drops to 59% at max.

How this story unfolded

10 days · 5 reports · 24 community posts · from Sep 22

  1. Sep 22
  2. Sep 23
  3. Sep 24
  4. Sep 25
  5. Sep 26
  6. Sep 27
  7. Sep 28
  8. Sep 29
  9. Sep 30
  10. Oct 1
  11. Oct 2

More stories today

Open the live feed