OpenAI has released GPT-6 Astra, a new model available to ChatGPT Plus, Pro, Business, and Enterprise users, as well as via the OpenAI API and AWS. The model is priced at $10/million input and $50/million output tokens, positioning it as a competitor to Claude Fable 5.
- GPT-6 Astra scores 99.9% on the ARC-AGI 3 benchmark using OpenAI's custom Provider Adapter harness, though the default harness yielded 62.7%.
- The model achieves 100% on ExploitBench and 99.2% on SRE-Bench binary reverse engineering, outperforming GPT-5.6 Sol in security tasks.
- It handles long context effectively, scoring 100% on the eight-needle benchmark at 256K–512K tokens and 96.3% at 512K–1M tokens.
- Artificial Analysis notes that Astra trails Claude Fable 5.1 in their Intelligence Index but leads the Coding Agent Index in cost efficiency, costing less than half of Fable for the same score.
The API model label will be gpt-6-astra upon full rollout.