GPT-6 Astra sets new NYT Connections benchmark record at 98.1 while costing ~40% less
zero0_one1 · reddit · 2026-09-05
A Reddit user reports new high scores on the NYT Connections LLM benchmark: GPT-6 Astra (xhigh reasoning) scores 98.1, with the high setting at 97.7 — both beating GPT-5.6 Sol while costing roughly 40% less per puzzle. The open-source benchmark lives at github.com/lechmazur/nyt-connections.
More from Models
- Fable 5.5 delayed: 'Stoptober' meme says no one dares ship a SOTA model now — patience_cave · 2026-10-21
- Decision Models' routing classifier hits AntSeed, cutting inference cost to 24% of always-Opus — airesearch12 · 2026-10-21
- Developer flags anomalous Claude Projects usage: 98% weekly limit consumed at 1% thread share — 0xkarasy · 2026-10-20
- DeepSeek hands inference providers 90%+ margins, pressuring Moonshot and Zhipu hosting deals — teortaxesTex · 2026-10-20
- HelixWorld world model tested: 70GB VRAM per BF16 GPU run, consumer cards need not apply — cocktailpeanut · 2026-10-20
- Code-only Paris at golden hour: Claude's render embarrasses ChatGPT in same-prompt test — EXM7777 · 2026-10-20