GPT-6 Astra Benchmark Details

GPT-6 Astra currently shows benchmark results led by GPQA Diamond (1 / 271, score 96), ARC-AGI-1 (1 / 91, score 98.50), ARC-AGI-2 (1 / 85, score 95). This page also tracks comparisons against 3 predecessor or same-series models.

Benchmark Results

GPT-6 Astra

Benchmark Results

General Knowledge

8 evaluations

Benchmark / mode

Score

Rank/total

ARC-AGI-1

Max

98.50

1 / 91

ARC-AGI-2

Max

1 / 85

ARC-AGI-3

Max

62.70

1 / 16

AA Intelligence Index

Max

61.20

7 / 28

HLE

MaxTools

57.20

12 / 190

Computer-use safety

MaxTools

2.40

1 / 1

Computer-use safety (with AutoReview)

MaxTools

1.80

1 / 1

Circumvention

MaxTools

1 / 1

Other

5 evaluations

Benchmark / mode

Score

Rank/total

GPQA Diamond

Max

1 / 271

HealthBench Professional (length-adjusted)

Max

63.40

1 / 1

LifeSciBench

Max

60.30

1 / 1

MedChemBench Internal

Max

49.30

1 / 1

GeneBench Pro

Max

37.80

1 / 1

AI Agent - Information Search

1 evaluations

Benchmark / mode

Score

Rank/total

BrowseComp

MaxTools

91.50

1 / 56

Coding and Software Engineer

7 evaluations

Benchmark / mode

Score

Rank/total

SRE-Bench (single-attempt)

MaxTools

1 / 1

SEC-Bench Pro

MaxTools

85.40

1 / 1

DeepSWE

MaxTools

74.10

2 / 35

AA Coding Agent Index

MaxTools

4 / 4

FrontierCode 1.1

MaxTools

64.50

1 / 6

Internal Database Migration

MaxTools

63.90

1 / 1

FrontierCode 1.1 Main

MaxTools

53.30

1 / 1

AI Agent - Tool Usage

10 evaluations

Benchmark / mode

Score

Rank/total

ExploitBench

MaxTools

100

1 / 2

Impossible ExploitGym

MaxTools

100

1 / 1

ScreenSpot-Pro (no tools)

Max

92.70

1 / 1

OSWorld 2.0

MaxTools

72.60

2 / 9

Terminal-Bench-Science 0.1

MaxTools

64.60

1 / 11

Terminal-Bench 4.0

MaxTools

57.90

1 / 13

ExploitGym (unlimited)

MaxTools

42.40

1 / 1

AutomationBench

MaxTools

41.40

5 / 14

ExploitBench (Jun–Aug 2026)

MaxTools

1 / 1

ExploitGym honeypot

MaxTools

1 / 1

Agent Level Benchmark

1 evaluations

Benchmark / mode

Score

Rank/total

Agents' Last Exam

MaxTools

59.30

1 / 15

Math and Reasoning

1 evaluations

Benchmark / mode

Score

Rank/total

FrontierMath Tier 4 v2

Max

97.60

1 / 41

Productivity Knowledge

3 evaluations

Benchmark / mode

Score

Rank/total

BenchCAD

MaxTools

95.90

1 / 1

Internal Design Tasks

MaxTools

1 / 1

Internal Data Science Tasks

MaxTools

40.90

1 / 1

Writing and Creative Capabilities

1 evaluations

Benchmark / mode

Score

Rank/total

OpenScore String Quartets

Max

0.84

1 / 1

Other

1 evaluations

Benchmark / mode

Score

Rank/total

Hallucination rate

Max

4.20

1 / 1

Other

2 evaluations

Benchmark / mode

Score

Rank/total

OpenAI MRCR v2 (8-needle, 256K–512K)

Max

100

1 / 1

OpenAI MRCR v2 (8-needle, 512K–1M)

Max

96.30

1 / 1

Compare with other models

Version History

How each version of the GPT-6 Astra series stacks up on benchmark tests

GPT-6 AstraGPT-5.6 Sol GPT-5.5 GPT-5.4

Benchmark categories:

The chart shows each model’s highest score per benchmark within the current filter. Out-of-100 benchmarks use raw heights; out-of-range benchmarks are scaled within that benchmark while labels keep the original scores.

12 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.· Click a row to view its trend chart.

Benchmark	GPT-6 AstraCurrent	GPT-5.6 Sol	GPT-5.5	GPT-5.4
AA Intelligence Index 综合评估	61.20Thinking Level · High	61.00Thinking Level · High ｜ Tools	--	--
ARC-AGI-1 综合评估	98.50Thinking Level · High	97.50Thinking Level · Extra High	95.00Thinking Level · Extra High	93.70Thinking Level · Extra High
ARC-AGI-2 综合评估	95.00Thinking Level · High	92.50Thinking Level · High	85.00Thinking Level · Extra High	77.10Standard Mode
ARC-AGI-3 综合评估	62.70Thinking Level · High	7.80Thinking Level · High	0.00Thinking Level · High	0.00Thinking Level · High
HLE 综合评估	57.20Thinking Level · High ｜ Tools	49.50Thinking Level · High	52.20Thinking Level · High ｜ Tools	52.10Thinking Level · Extra High ｜ Tools
GPQA Diamond 科学与综合推理	96.00Thinking Level · High	93.50Thinking Level · High	94.00Thinking Level · Extra High	92.80Thinking Level · Extra High
BrowseComp AI Agent - 信息收集	91.50Thinking Level · High ｜ Tools	--	84.40Thinking Level · High ｜ Tools	82.70Thinking Level · Extra High ｜ Tools
AA Coding Agent Index 编程与软件工程	67.00Thinking Level · High ｜ Tools	80.00Thinking Level · Extra High ｜ Tools	--	--
DeepSWE 编程与软件工程	74.10Thinking Level · High ｜ Tools	72.70Thinking Level · Extra High ｜ Tools	67.00Thinking Level · Extra High ｜ Tools	52.00Thinking Level · Extra High ｜ Tools
FrontierCode 1.1 编程与软件工程	64.50Thinking Level · High ｜ Tools	60.60Thinking Level · High ｜ Tools	--	--
OSWorld 2.0 AI Agent - 工具使用	72.60Thinking Level · High ｜ Tools	62.60Thinking Level · Extra High ｜ Tools	--	--
Terminal-Bench 4.0 AI Agent - 工具使用	57.90Thinking Level · High ｜ Tools	37.27Thinking Level · High ｜ Tools	--	--

3 additional benchmarks remain in the chart above.

Single-Benchmark Version Trend

Viewing: AA Intelligence Index · 综合评估

Benchmark

NormalNormal + ToolsThinkingThinking + ToolsDeepDeep + Tools

X-axis shows model and release date, Y-axis shows score; solid lines connect the same mode across versions, while dotted guides align modes within the same generation.

Standard API Pricing Across the GPT-6 Astra Series

Shows standard text input and output pricing side by side for each model. If extended-context pricing exists, the chart keeps the base rate and explains the threshold below.

Source: DataLearnerAI. Standard text prices shown here use the default supplier. · USD / 1M tokens

When a context threshold exists, the charted base price only applies within these limits:

GPT-6 Astra: Base price applies to <= 272000

Model	Supplier	Standard input	Standard output	Base price applies to
GPT-6 Astra	OpenAI	$10 / 1M tokens	$50 / 1M tokens	<= 272000
GPT-5.6 Sol	OpenAI	$4 / 1M tokens	$20 / 1M tokens	—
GPT-5.5	OpenAI	$5 / 1M tokens	$30 / 1M tokens	—
GPT-5.4	OpenAI	$2.5 / 1M tokens	$15 / 1M tokens	—