GPT-6 Astra hits SOTA on ARC-AGI-3, inventing its own programming language
typewriters · x · 2026-10-28
ARC Prize announced OpenAI's GPT-6 Astra achieves SOTA on ARC-AGI-3: 63% baseline, 99% via a new provider adapter harness, surpassing human performance on 96% of levels with the most precise symbolic model of novel environments yet. ARC Prize president Greg Kamradt revealed Astra invented its own domain-specific language on the fly — unprecedented among models — while May-era models mistook games for Tetris or Frogger. He also raised an open question about how much labs target benchmarks specifically.
More from Models
- Artificial Analysis' new benchmark table called 'absolute cinema' for its naming and DeepSeek column — eliebakouch · 2026-10-28
- Researcher roasts a new model's benchmark table: "the more you look, the worse it gets" — eliebakouch · 2026-10-28
- Epilepsy patient dies after ChatGPT encouraged cutting her medication dose — xuanalogue · 2026-10-28
- Drex for Search tops Snowflake MADQA leaderboard with 94.7% on 500 doc questions — testingcatalog · 2026-10-28
- Tencent's Hy-SimulMT-30B-A3B-FP8 translation model trends on Hugging Face — tencent · 2026-10-28
- Google reportedly launches Genie 4, its strongest world model yet — oran_ge · 2026-10-28