Tihu · Pelican Bench

GPT-6 Astra

OpenAI · Model release:

AI reference score 8.81 ± 0.19 · max (best) Last 4 weekly runs · 12 images
Tier comparison
TierSamplesAI reference scoreValid SVG rateMean timeSamples
max
GPT-6 Astra · max · 2026-10-26 · #1 · 8.81
 8.81 ± 0.19Last 4 weekly runs · 12 images 100%365 s3
medium default
GPT-6 Astra · medium · 2026-10-26 · #2 · 8.43
 8.20 ± 0.54Last 4 weekly runs · 12 images 100%66 s3
low
GPT-6 Astra · low · 2026-10-26 · #1 · 8.55
 7.94 ± 0.64Last 4 weekly runs · 12 images 100%79 s3
GPT-6 Astra · max · 2026-10-26 · #1 · 8.81
max 8.81 ± 0.19Last 4 weekly runs · 12 images
Valid SVG rate 100% · Mean time 365 s
GPT-6 Astra · medium · 2026-10-26 · #2 · 8.43
medium default8.20 ± 0.54Last 4 weekly runs · 12 images
Valid SVG rate 100% · Mean time 66 s
GPT-6 Astra · low · 2026-10-26 · #1 · 8.55
low 7.94 ± 0.64Last 4 weekly runs · 12 images
Valid SVG rate 100% · Mean time 79 s

Latest samples

max

AI reference score 8.81 ± 0.19 Last 4 weekly runs · 12 images Normal
Top tiers of one modelView tier page

MonLatest

GPT-6 Astra · max · 2026-10-26 · #1 · 8.81GPT-6 Astra · max · 2026-10-26 · #2 · 8.78GPT-6 Astra · max · 2026-10-26 · #3 · 8.78

medium

AI reference score 8.20 ± 0.54 Last 4 weekly runs · 12 images Normal
Top tiers of one modelView tier page

MonLatest

GPT-6 Astra · medium · 2026-10-26 · #2 · 8.43GPT-6 Astra · medium · 2026-10-26 · #3 · 8.35GPT-6 Astra · medium · 2026-10-26 · #1 · 7.22

low

AI reference score 7.94 ± 0.64 Last 4 weekly runs · 12 images Normal
Top tiers of one modelView tier page

MonLatest

GPT-6 Astra · low · 2026-10-26 · #1 · 8.55GPT-6 Astra · low · 2026-10-26 · #2 · 8.19GPT-6 Astra · low · 2026-10-26 · #3 · 7.84

Score trend

max Normal

2026-10-08 · 8.842026-10-09 · 8.672026-10-10 · 8.552026-10-11 · 8.672026-10-12 · 8.752026-10-13 · 8.882026-10-14 · 8.772026-10-15 · 8.792026-10-16 · 8.772026-10-17 · 8.852026-10-18 · 8.792026-10-19 · 8.672026-10-20 · 8.722026-10-21 · 8.532026-10-22 · 8.482026-10-23 · 8.822026-10-24 · 8.852026-10-25 · 8.792026-10-26 · 8.79

19 records in the last 30 days

2026-10-26 · Normal2026-10-25 · Normal2026-10-24 · Normal2026-10-23 · Normal2026-10-22 · Normal2026-10-21 · Normal2026-10-20 · Normal2026-10-19 · Normal2026-10-18 · Normal2026-10-17 · Normal2026-10-16 · Normal2026-10-15 · Normal2026-10-14 · Normal2026-10-13 · Collecting earlier results2026-10-12 · Collecting earlier results2026-10-11 · Collecting earlier results2026-10-10 · Collecting earlier results2026-10-09 · Collecting earlier results2026-10-08 · Collecting earlier results

medium Normal

2026-10-08 · 7.672026-10-09 · 7.482026-10-10 · 7.482026-10-11 · 8.172026-10-12 · 8.362026-10-13 · 8.332026-10-14 · 8.222026-10-15 · 8.422026-10-16 · 8.422026-10-17 · 8.352026-10-18 · 8.532026-10-19 · 8.012026-10-20 · 7.662026-10-21 · 8.372026-10-22 · 7.672026-10-23 · 8.402026-10-24 · 8.202026-10-25 · 8.222026-10-26 · 8.00

19 records in the last 30 days

2026-10-26 · Normal2026-10-25 · Normal2026-10-24 · Normal2026-10-23 · Normal2026-10-22 · Normal2026-10-21 · Normal2026-10-20 · Normal2026-10-19 · Normal2026-10-18 · Normal2026-10-17 · Normal2026-10-16 · Normal2026-10-15 · Normal2026-10-14 · Normal2026-10-13 · Collecting earlier results2026-10-12 · Collecting earlier results2026-10-11 · Collecting earlier results2026-10-10 · Collecting earlier results2026-10-09 · Collecting earlier results2026-10-08 · Collecting earlier results

low Normal

2026-10-08 · 7.312026-10-09 · 8.282026-10-10 · 7.042026-10-11 · 7.972026-10-12 · 8.482026-10-13 · 7.982026-10-14 · 7.802026-10-15 · 8.572026-10-16 · 8.132026-10-17 · 8.352026-10-18 · 8.642026-10-19 · 7.562026-10-20 · 7.302026-10-21 · 7.802026-10-22 · 6.992026-10-23 · 7.042026-10-24 · 8.312026-10-25 · 8.242026-10-26 · 8.19

19 records in the last 30 days

2026-10-26 · Normal2026-10-25 · Normal2026-10-24 · Normal2026-10-23 · Normal2026-10-22 · Normal2026-10-21 · Normal2026-10-20 · Normal2026-10-19 · Normal2026-10-18 · Normal2026-10-17 · Normal2026-10-16 · Normal2026-10-15 · Normal2026-10-14 · Normal2026-10-13 · Collecting earlier results2026-10-12 · Collecting earlier results2026-10-11 · Collecting earlier results2026-10-10 · Collecting earlier results2026-10-09 · Collecting earlier results2026-10-08 · Collecting earlier results

Earlier samples

max

medium

low

Versions by score

  1. GPT-6 Astra
    GPT-6 Astra · max
    max 8.81Last 4 weekly runs · 12 images

Scores by criterion

Scores by tier
Scoring criteria max medium low
Pelican likeness9.008.758.58
Bicycle correctness8.507.508.00
Riding pose8.757.758.17
Composition9.088.508.50
Obvious errors (higher is better)8.677.427.58

Release notes

Original (Chinese)当天向有限组织开始推出；随后数日向 ChatGPT Plus/Pro/Business/Enterprise 以及 OpenAI API、Azure、AWS Bedrock 开放。

· View release source ↗