What GPT-6 Astra's 99.9% ARC-AGI-3 Score Actually Measures
mixtapedmonk · reddit · 2026-09-18
Digging into ARC Prize's actual results table, the author found GPT-6 Astra's headline 99.9% ARC-AGI-3 score hides a 37-point gap between two harnesses on the same model — and the test's own creators won't call it AGI. Fortune reported five numbers quietly changed on OpenAI's launch page after going live. The writeup includes the full harness breakdown and the Llama 4 precedent, with all official sources linked.
More from Models
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02
- OpenAI Dots user says bloat and broken infra drove him to try newcomer Muse — Koala_Confused · 2026-11-02
- Fable 5.5 reportedly one-shots animation in pure code as Anthropic's coding bet pays off — Hesamation · 2026-11-02
- One agent goal has been running for 4 days, burning 30K tokens — CtrlAltDwayne · 2026-11-02
- Daily AI Brief: Gemini 4 Carbon spotted on Arena, DeepSeek open-sources coding agent — testingcatalog · 2026-11-02