OpenAI’s GPT-6.2 Astra model has achieved a score of 174 on the Epoch Capabilities Index, a composite benchmark managed by Epoch AI. This result aligns the model with Anthropic’s Claude Fable 5.2, which recorded an identical score shortly after its release on November 9, 2026.
The evaluation for GPT-6.2 Astra was published by Epoch AI on November 17, 2026. Both models currently exceed the established pre-Astra trend line for reasoning AI systems. The index itself aggregates over 50 individual benchmarks across various disciplines into a single metric.
The new score provides a comparative baseline against previous models. For context, the index anchors GPT-5 at 150 points and Claude 3.5 Sonnet at 130 points. Consequently, both the latest GPT and Claude iterations sit 24 points above the GPT-5 marker and 44 points above Claude 3.5 Sonnet.
The performance of GPT-6.2 Astra represents a shift from the previous trajectory of the GPT-6 Astra series. Between September and early November 2026, earlier models in this line consistently scored between 167 and 169. The latest result marks a departure from that incremental range.
Epoch AI emphasizes that the updated scores are derived exclusively from third-party evaluations. No self-reported benchmark data from either OpenAI or Anthropic was included in the current assessment of the models.






