HLE gains appear to be slowing as GPT-5.6 and Claude 5 add only a few points
thegamebegins25 · reddit · 2026-07-25
A Reddit post argues that progress on Humanity’s Last Exam (HLE) may be slowing.
The author points out that:
- OpenAI’s GPT-5.6 blog post did not mention HLE;
- GPT-5.6 Sol (max) seems to beat GPT-5.5 (xhigh) by less than 3 points;
- Claude’s newest Fable/Opus/Sonnet 5 line also appears to improve by only about 3 points over Opus 4.8.
The post contrasts this with last year’s more frequent double-digit monthly gains and asks whether companies are now shifting focus to Agent’s Last Exam because HLE is getting harder to move.
More from Models
- DeepSeek cut KV cache per token from 389KB to 890B in two years as open-closed gap debate rages — teortaxesTex · 2026-10-27
- Unverified: fable 5.5 speculated at 4-5T parameters to outscale 1-3T Opus — haider1 · 2026-10-27
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27
- Custom VQA benchmark under 48GB VRAM: Qwen3.8-Flash-Next leads, Qwen3.8-27B regressed — EmilPi · 2026-10-27