OpenAIUnited States· released Sep 2026

GPT-6 Astra

Name: GPT-6 Astra
Author: OpenAI

Top of the OpenCharts Index, with the best published result on 9 of the 14 tests it has taken.

OpenAI, United States. One of 26 models from this lab on the Index.

Third-party publishedMedium confidence (7+ comparable tests, 3+ Index categories)

Available in Theo as Theo Explore

Theo orchestrates it for the steps it does best and always shows which engine answered.

Use it in Theo

Snapshot September 8, 2026

OpenCharts Index

99.0

Ranked #1 of 36 · 1.0% behind the frontier on average · 14 comparable tests · 3 counting categories

Drop any single test and the Index lands between 99.0 and 99.7.

Needs more results (3 of 8 so far)

Averaging 100.0 on the 3 it has; the score opens at 4 of 8.

Distance to today's frontier on the hardest tests. Not a measure of intelligence.

What it's known for

GPT-6 Astra, in 14 results.

Strongest in math (99.8) and reasoning (99.6). It holds the best published result on ARC-AGI-1, ARC-AGI-2 and Code Arena (WebDev) and 6 more. It trails the frontier most in coding, 2.3% behind on average. No comparable results yet in multimodal, human preference and long context. Within 5% of the best published results, on average.

Best published result onARC-AGI-198.5%ARC-AGI-295.0%Code Arena (WebDev)1,797 FrontierMath (Tiers 1–3)93.7%FrontierMath Tier 497.6%GPQA Diamond95.8%Mock AIME 2024–2025100.0%SimpleQA Verified75.6%WeirdML92.9%

Strongest category

Math

Competition and research mathematics, graded on the final answer or the proof.

99.8

Mock AIME 2024–2025100.0% · best
FrontierMath (Tiers 1–3)93.7% · best
FrontierMath Tier 497.6% · best
ProofBench99.0% · 99% of best

Mean of 4 of the category's 5 comparable tests. Counts toward the Index.

Best single result

ARC-AGI-1

Solve the original Abstraction and Reasoning Corpus: small grids, a handful of examples, one hidden rule.

98.5%

Best published result of 51 models.

The first ARC set is close to saturated at the frontier, so it now shows whether smaller models can generalize at all.

Where it trails

SciCode

Write research-grade scientific code, one sub-problem at a time, that passes unit tests.

56.5%

9.0% behind the best published result (62.0%) · #9 of 73.

Scientific programming needs the math and the code to be right at the same time.

Not measured yet

No comparable result yet in these 3 categories. A blank is a blank, never a zero, and it does not lower the Index.

Multimodal
Human preference
Long context

Covered, but not averaged into the Index:

Knowledgesingle-test category, shown beside the Index · 100.0
Agentic1 of 8 tests · needs 4 · mean 98.5

Every score, with its source

See it at work

What GPT-6 Astra was asked to do, and how it did.

Every test it has taken, by category, with the kind of task it faced, who produced the number and where the result landed against the best published one. The examples are original and illustrative, never items from the datasets themselves.

Knowledge · mean 100.0 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

SimpleQA VerifiedEpoch-run

Answer a short factual question exactly, without searching, or say you do not know.

75.6%

best published result
of 57 models

max

Source: Epoch AI (internal runs)

Every result is read against the best published one on its test: percent scores as a share of the best above chance, Arena ratings as the expected win rate against the board leader, open-ended values on a log scale. Sitting a harder exam never lowers a model. Every number links to the publisher that produced it.

Head to head

GPT-6 Astra against whoever you pick.

Choose a rival. Every counted test both have taken appears side by side as a share of the best published result on that test, with the category means above. A win is a higher share; a gap under 2 points is a tie, because one item on a 45-problem exam is 2.2 points and publishers' own standard errors are of that order.

GPT-6 Astra

#1·Index 99.0A+

6 : 1

wins · 6 ties · 13 shared

Claude Fable 5.1

#2·Index 97.0A+

GPT-6 Astra comes out ahead on 6 of the 13 tests both have taken.

Category means

Reasoning
99.6
97.5
Knowledge
100.0
93.7
Coding
97.7
97.0
Math
99.8
96.6
Agentic
98.5
88.4
Human preference
—
99.2

Shared tests · biggest gaps first

Code Arena (WebDev)Coding
1,797
1,762
+10.0
FrontierMath Tier 4Math
97.6%
87.8%
+10.0
SciCodeCoding
56.5%
62.0%
−9.0
SimpleQA VerifiedKnowledge
75.6%
70.8%
+6.3
ARC-AGI-2Reasoning
95.0%
90.0%
+5.3
FrontierCodeCoding
53.3%
50.9%
+4.4
FrontierMath (Tiers 1–3)Math
93.7%
90.2%
+3.7
CritPtReasoning
31.7%
31.1%
tie
APEX-AgentsAgentic
46.7%
47.4%
tie
ARC-AGI-1Reasoning
98.5%
97.5%
tie
ProofBenchMath
99.0%
100.0%
tie
WeirdMLCoding
92.9%
92.9%
tie
Mock AIME 2024–2025Math
100.0%
100.0%
tie

Bars are distances to the best published result on each test (percent scores above chance, Arena ratings as win rate against the leader, open-ended values on a log scale). Raw values as the publishers report them.

Open Claude Fable 5.1

Scores

Every test, every source

The best published run per benchmark, the raw value as the publisher reports it, who produced the number, how close it comes to the best published result on that test, and where the model sits among every model scored on it. Percent scores are read above chance, Arena ratings as a win rate against the board leader, open-ended values on a log scale. A test with too few models to compare against is shown but not counted.

Category means

Reasoning99.6 · 4/6
Knowledgebeside100.0 · 1/1
Coding97.7 · 4/7
Math99.8 · 4/5
Agentic98.5 · 1/8
Multimodal—
Human preference—
Long context—

Mean · tests taken of the category's comparable tests. A faded category either rests on fewer than half of them or is a single-test category (marked “beside”), which is shown next to the Index and never averaged into it.

GPQA DiamondEpoch-runReasoning

95.8%

best published result
of 78 models

max

Source: Epoch AI — AI Benchmarking Hub

ARC-AGI-2LeaderboardReasoning

95.0%

best published result
of 51 models

max

Source: Epoch AI — AI Benchmarking Hub

ARC-AGI-1LeaderboardReasoning

98.5%