GPT-6 Astra scores 62.7% on ARC-AGI-3 with the standard harness and 99.9% with a new provider adapter harness; Claude Opus 5 scored 30.2%, and GPT-5.6 Sol 7.8% (Greg Kamradt/ARC Prize)
Astra’s ARC-AGI-3 result changes sharply depending on the harness: 62.7% standard, 99.9% with a new adapter.
ARC Prize says the standard setup lets the model decide what notes to carry forward, while the provider adapter preserves opaque reasoning between requests and uses compaction. The group says it will test new ARC-AGI-3 models with both harnesses going forward. Claude Opus 5 scored 30.2%, and GPT-5.6 Sol scored 7.8% in the cited comparison. Techmeme's note
ARC Prize says the standard setup lets the model decide what notes to carry forward, while the provider adapter preserves opaque reasoning between requests and uses compaction. The group says it will test new ARC-AGI-3 models with both harnesses going forward. Claude Opus 5 scored 30.2%, and GPT-5.6 Sol scored 7.8% in the cited comparison. Techmeme's note
score 8