Enter opens the highlighted result. Esc closes.

Initial Stage leaderboard

As of November 22, 2026, among the results listed here, OpenAI GPT-6.2 Astra has the highest reported Initial Stage score, 81.8% (mean rubric score), published by OpenAI, ahead of OpenAI GPT-6 Astra at 78.0%. 5 models have a Initial Stage result in our records.

Scores as the model makers published them, not re-run by us; each row links to its announcement.

Best reported score

81.8%

OpenAI GPT-6.2 Astra

Reported by OpenAI, November 17, 2026 · mean rubric score

Runner-up

78.0%

OpenAI GPT-6 Astra

Reported by OpenAI, November 17, 2026 · mean rubric score

5 models ranked · 1 maker

Reported Initial Stage scores

Initial Stage scores reported by model makers
Rank Model Score Setting Reported by Date
1 GPT-6.2 Astra 81.8% mean rubric score OpenAI November 17, 2026 (model release)
2 GPT-6 Astra 78.0% mean rubric score OpenAI November 17, 2026 (model release)
3 GPT-6.1 Terra 76.4% mean rubric score OpenAI November 17, 2026 (model release)
4 GPT-6.2 Sol 74.7% mean rubric score OpenAI November 17, 2026 (model release)
5 GPT-5.6 Sol 68.3% mean rubric score OpenAI November 17, 2026 (model release)

Other reported runs

Initial Stage scores reported by model makers, other reported runs
Model Score Setting Reported by Date
GPT-5.6 Sol 61.1% mean rubric score OpenAI November 17, 2026 (model release)
GPT-6 Astra 66.7% mean rubric score OpenAI November 17, 2026 (model release)
GPT-6.1 Terra 45.3% mean rubric score OpenAI November 17, 2026 (model release)
GPT-6.2 Astra 76.9% mean rubric score OpenAI November 17, 2026 (model release)
GPT-6.2 Sol 74.7% mean rubric score OpenAI November 17, 2026 (model release)

How much to trust these numbers

Source
Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
Settings
All ranked results use the same reported setting.
Spread
81.8% to 68.3% across 5 models.
Freshness
Published November 17, 2026.

What Initial Stage measures

Initial Stage appears in the release announcements of OpenAI. We have not found a public write-up of the benchmark itself yet.