Tihu · Pelican Bench

GPT-6.1 Terra

OpenAI · Model release:

AI reference score 6.22 ± 0.85 Last 2 weekly runs · 6 images 2 weekly tests leftUpdated every Sunday · 2026-11-01Top tiers of one model Compare with previous version

Latest samples

SunLatest

GPT-6.1 Terra · low · 2026-11-01 · #2 · 7.73GPT-6.1 Terra · low · 2026-11-01 · #3 · 5.66GPT-6.1 Terra · low · 2026-11-01 · #1 · 5.29

Score trend

low Collecting earlier results

2026-10-26 · 6.222026-11-01 · 6.23

2 records in the last 30 days

2026-11-01 · Collecting earlier results2026-10-26 · Collecting earlier results

Earlier samples

low

Mon

GPT-6.1 Terra · low · 2026-10-26 · #2 · 6.52GPT-6.1 Terra · low · 2026-10-26 · #3 · 6.17GPT-6.1 Terra · low · 2026-10-26 · #1 · 5.95

Versions by score

  1. GPT-6.1 Terra
    GPT-6.1 Terra · low
    low 6.22Last 2 weekly runs · 6 images
  2. GPT-5.6 Terra
    GPT-5.6 Terra · low
    low 5.91Last 4 weekly runs · 12 images

Scores by criterion

Pelican likeness
7.67
Bicycle correctness
5.92
Riding pose
4.83
Composition
6.92
Obvious errors (higher is better)
5.50

Release notes

Original (Chinese)当天向 ChatGPT Work（Plus/Pro/Business/Enterprise/Edu）、Codex 及 API（gpt-6.1-terra）开放；公告写明 Chat 暂未上线。

· View release source ↗