OpenAI's GPT-6 Astra on ARC-AGI-3 Scored 99.95% with Provider Adapter Harness

OpenAI·Sep 2, 2026·12 harness configurations

GPT-6 Astra's best observed ARC-AGI-3 Semi-Private result with the Standard harnessStandard harness enables a model to carry forward notes it chooses to keep with it throughout the environment. was 62.7% at max reasoning for $26,098. With the Provider Adapter harnessThe Provider Adapter harness preserves opaque reasoning state between requests and uses compaction for longer conversations, allowing the model to reuse prior work., its best observed result was 99.9% at high reasoning for $18,817. Read more in our blog post, OpenAI's GPT-6 Astra on ARC-AGI-3.

ARC-AGI 3 leaderboard #

GPT-6 AstraGPT-6 Astra - Provider Adapter

Verified scores #

| Variant | ARC-AGI-1 | ARC-AGI-2 | ARC-AGI-3(Standard harness) | ARC-AGI-3(Provider Adapter harness) | 
|---|---|---|---|---|

| Max | 97.5% | 95.0% | 62.71% | 98.55% | | XHigh | 98.5% | 93.3% | 59.34% | 98.44% | | High | 98.5% | 92.1% | 54.82% | 99.95% | | Medium | 97.5% | 92.1% | 38.59% | 98.44% | | Low | 96.5% | 85.4% | 17.45% | 98.03% | | None | 86.0% | 59.6% | 35.18% | 96.72% |

Tasks & environments #

Pass/fail per reasoning level across each benchmark.

ARC-AGI-3 Public Demo - Standard harnessStandard harness enables a model to carry forward notes it chooses to keep with it throughout the environment.

25 environments

ARC-AGI-3 Public Demo - Provider Adapter harnessThe Provider Adapter harness preserves opaque reasoning state between requests and uses compaction for longer conversations, allowing the model to reuse prior work.

25 environments

ARC-AGI-2 Public Eval

120 tasks

ARC-AGI-1 Public Eval

400 tasks