GPT-6 Astra hits record 169 on Epoch AI's ECI, sweeping math and continual-learning benchmarks
rohanpaul_ai · x · 2026-09-05
Epoch AI reports GPT-6 Astra scored a record 169 on its ECI capability index, up 6 points from the previous best of 163, while also setting new records on its math, continual learning, and game-puzzles benchmarks.
- ECI aggregates 50+ benchmarks into one capability scale, reducing reliance on any single saturated test
- Epoch estimates benchmark difficulty from overlapping model results, so hard-test scores carry more information
- The jump is large but within Epoch's uncertainty range for the reasoning-era ECI trend
- On the long-horizon coding benchmark MirrorCode, Astra ranks between Opus 4.7 and Fable 5
- OpenAI granted Epoch pre-release access; charts are in the thread
Related event: GPT-6 Astra Sets Epoch AI ECI Record at 169(3 posts)→
More from Models
- Diffusion LLM bias rankings shift with sampling configs, persist after debiasing — Bollegala · 2026-10-23
- Running ~100 DeepSeek v4.1 flash agents in parallel for under $20/month — zack_overflow · 2026-10-23
- Researcher laments open-source audio models are surprisingly subpar, even representation ones — gowthami_s · 2026-10-23
- User claims they generated video with "Claude Opus 5.5", unannounced model name surfaces — jwt0625 · 2026-10-23
- Security researchers push back as OpenAI restricts Astra access; BlackHat 2027 mocked as 'cyber model' bazaar — cyb3rops · 2026-10-23
- Mistral Large 4 Leaked on Hugging Face: 1T Params, 49B Active, Open Weights Oct 31 — mervenoyann · 2026-10-23