Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT-6 Astra

OpenAI

Strongest in Data analysis (#1 of 57), weakest in Safety (#158 of 337). Above par in 23 of 24 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Opus 5.5 and ahead of Claude Opus 4.8 and GLM-5.3.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Core abilities67.1#2/204
Data analysis66.1#1/57
Language69#3/57
General intelligence68.5#4/204
Instruction following63.5#7/57
Reasoning66.4#2/178
Reasoning68.6#2/140
Mathematics65.3#3/140
Coding62.7#5/165
Agentic coding64.7#5/157
Code generation56.1#22/77
Agents63.4#12/268
Knowledge work67.5#12/178
Human preference64.5#23/342
Human preference64.5#23/342
Professional56#28/168
IT operations73.8#1/11
Biology research68.5#2/16
Cybersecurity27.4#9/9
Medical58.6#16/140
Legal53.5#49/151
Finance54.5#52/151
Safety51#158/337
Factual grounding53.2#43/101

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for GPT-6 Astra, left for the other.

GLM-5won 15 · lost 0
Mimo V2.5 Prowon 15 · lost 0
GLM-4.6won 16 · lost 0
Mistral Large 3won 17 · lost 0
Kimi K2.5won 17 · lost 0
Qwen3.6 Pluswon 19 · lost 0
Gemini 3.5 Flash Litewon 19 · lost 0
Grok 4.3won 20 · lost 0
DeepSeek V4 Flashwon 21 · lost 1
GPT-6 Lunawon 20 · lost 1
Kimi K2.6won 20 · lost 1
MiniMax M3won 19 · lost 1
Qwen3.7 Maxwon 19 · lost 1
Gemini 3.5 Flashwon 19 · lost 1
Qwen3.8 27Bwon 18 · lost 1
DeepSeek V3won 17 · lost 1
Claude Sonnet 4won 17 · lost 1
DeepSeek R1won 17 · lost 1
GPT OSS 120Bwon 16 · lost 1
Gemini 2.5 Flashwon 16 · lost 1
Claude Haiku 4.5won 16 · lost 1
DeepSeek V3.2won 16 · lost 1
Qwen3.6 27Bwon 15 · lost 1
GPT-4o Miniwon 15 · lost 1
GPT-4.1 Nanowon 15 · lost 1
GPT-4.1 Miniwon 15 · lost 1
Kimi K2won 15 · lost 1
Qwen3 235B A22B Instructwon 15 · lost 1
Llama 4 Scout Instructwon 14 · lost 1
Command Awon 14 · lost 1
O3 Miniwon 14 · lost 1
Grok 4 Fastwon 14 · lost 1
Grok 4won 14 · lost 1
Gemma 4 31Bwon 14 · lost 1
GPT-5.4 Nanowon 19 · lost 2
Grok 4.5won 19 · lost 2
GLM-5.2won 19 · lost 2
Inklingwon 18 · lost 2
GPT-5.4 Miniwon 18 · lost 2
GPT-5.6 Terrawon 18 · lost 2
O4 Miniwon 15 · lost 2
Claude Sonnet 4.5won 15 · lost 2
Claude Opus 4won 14 · lost 2
Gemini 3.1 Flash Litewon 14 · lost 2
GPT-6 Solwon 19 · lost 3
DeepSeek V4.1 Flashwon 19 · lost 3
GPT-5.6 Lunawon 18 · lost 3
Claude Sonnet 4.6won 18 · lost 3
GPT-5.4won 18 · lost 3
Gemini 3.6 Flashwon 18 · lost 3
GLM-5.3 Flashwon 17 · lost 3
GPT-5 Miniwon 14 · lost 3
GPT-5won 14 · lost 3
DeepSeek V4 Prowon 18 · lost 4
O3won 13 · lost 3
Gemini 3.1 Prowon 17 · lost 4
GPT-5.2won 16 · lost 4
Claude Sonnet 5won 16 · lost 4
Claude Opus 4.5won 16 · lost 4
Gemini 2.5 Prowon 14 · lost 4
Gemini 3 Prowon 12 · lost 4
Grok 4.7won 15 · lost 6
Muse Spark 1.1won 14 · lost 6
Muse Spark 1.2won 14 · lost 6
GPT-4.1won 11 · lost 5
Grok 4.6won 15 · lost 7
GPT-5.5won 15 · lost 7
Gemini 3.8 Flashwon 15 · lost 7
GPT-5.1won 10 · lost 5
GPT-5.6 Solwon 16 · lost 8
Gemini 3.7 Flashwon 13 · lost 7
Claude Opus 4.6won 12 · lost 7
Kimi K3won 13 · lost 8
Claude Opus 4.7won 13 · lost 8
Qwen3.8 Maxwon 12 · lost 8
GLM-5.3won 12 · lost 8
Claude Opus 4.8won 12 · lost 8
Muse Spark 1.3won 9 · lost 10
Claude Opus 5won 10 · lost 13
Mimo V2.6 Prowon 6 · lost 9
Claude Fable 5.1won 6 · lost 15
Claude Opus 5.5won 5 · lost 17
Claude Fable 5won 4 · lost 15

§ 3 · Sources

Where the numbers come from

10 publications, 27 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1478
Artificial Analysis ↗1 measureread 2026-09-26
Artificial Analysis Intelligence Index 53
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 1542
AA-Briefcase ↗1 measureread 2026-09-26
AA-Briefcase 1569
Terminal-Bench ↗1 measureread 2026-09-26
Terminal-Bench 58.2% (Codex)
ARC-AGI-2 ↗1 measureread 2026-09-26
ARC-AGI-2 95%
LiveBench ↗8 measuresread 2026-09-26
LiveBench 82.2LiveBench · Reasoning 92.7LiveBench · Coding 80.4LiveBench · Agentic Coding 57.3LiveBench · Mathematics 96.8LiveBench · Data Analysis 83LiveBench · Language 89.4LiveBench · Instruction Following 75.6
SimpleBench ↗1 measureread 2026-09-26
SimpleBench 83.6%
Vals.ai ↗11 measuresread 2026-09-26
Vals · Legal Research Bench 39.42%Vals · Harvey Legal Agent Benchmark 5.42%Vals · Finance Agent 53.54%Vals · MedCode 48.49%Vals · MedScribe 87.91%Vals · BioMysteryBench 79.26%Vals · CyberBench 41.07%Vals · SRE Bench 56.87%Vals · Vibe Code Bench 89.59% (OpenHands)Vals · Code Migration 67.74%Vals · ProofBench 99%
Vectara · Factual consistency 91.3%

Badge

PublicAI Index badge for GPT-6 Astra[![PublicAI Index](https://publicai.io/model-index/badge?model=gpt-6-astra)](https://publicai.io/model-index/m/gpt-6-astra)