Astra-max hits 95% on eyebench-v3 at half the cost of Sol-max, tokens ~3.8x fewer
adonis_singh · x · 2026-09-05
A blogger reports Astra-max scoring 95% on the vision benchmark eyebench-v3, at roughly half the cost of Sol-max while outputting 3.8x fewer tokens and nearly doubling its score. He adds that on an intelligence-per-token basis it beats everything else even at low reasoning settings, and says he'll move on to harder benchmarks instead of making a v4. Note: unofficial third-party evaluation.
More from Models
- 105 Real Bugs Tested: GPT-6.1 Sol Over 10x Cheaper Than GPT-5.6 Sol — PawelHuryn · 2026-10-23
- Abliterated V4.1-Flash eval: abstract algebra drops 6pp but college math gains 5pp — teortaxesTex · 2026-10-23
- Claude users hit "model at capacity" errors hourly, canceling plans over it — Alex-Stack · 2026-10-23
- Kimi K3.1 launch imminent: 3T-param bet on RL seen as make-or-break — teortaxesTex · 2026-10-23
- Uncensored fine-tunes vanish from Hugging Face amid unexplained takedowns — umarmnaq · 2026-10-23
- Claude app menu leaks mystery model name 'Fable' as user hits limits — ezshine · 2026-10-23