Skip to content
Leaderboard / OpenAI

GPT-6 Astra

OpenAI Max CLOSED Proprietary
Verdict

GPT-6 Astra is a top-tier model: #8 of 406 by SIscope score. Strongest at reasoning & knowledge, weaker at chat quality. On the pricey side: costs more than 99% of priced models.

SIscope score
93
#8 of 406
Human-vote rank
#75
of 419 · score 1441
Price per 1M tokens, blended
$20
$10 in · $50 out
Context window
1.1M
1,050,000 tokens

Shape

Each spoke is a category, 0 at the centre, 100 at the edge.

CodingAgentsReasoningMathChatInstructions

Category scores

Percentiles: 90 means better than about 90% of tested models.

Coding95
Agents & tools95
Reasoning & knowledge95
Math94
Chat quality82
Following instructions89

What goes into each category · Compare with DeepSeek V4.1 Flash

Human-vote score over time

Text arena, recent updates.

143514431450Oct 20: 1441 (rank #73)Oct 22: 1442 (rank #73)Oct 28: 1441.2 (rank #75)Oct 20Oct 28

Coding votes over time

WebDev arena, recent updates.

176517751785Oct 16: 1777.2 (rank #2)Oct 18: 1775.7 (rank #2)Oct 19: 1773.8 (rank #2)Oct 22: 1774.7 (rank #2)Oct 23: 1774.3 (rank #2)Oct 27: 1773.5 (rank #2)Oct 28: 1773.4 (rank #2)Oct 16Oct 28

Independent test runs

Run by Epoch AI on the same harness for every model. Bars show the share of questions answered correctly.

GPQA Diamond96%
PhD-level science multiple choice questions
Better than 100% of 333 tested models. Run Aug 30, 2026.
SimpleQA Verified76%
Short factual questions; rewards not making things up
Better than 99% of 100 tested models. Run Aug 30, 2026.
OTIS Mock AIME 2024-25100%
Competition maths problems in the style of AIME
Better than 96% of 313 tested models. Run Aug 30, 2026.
FrontierMath Tiers 1-394%
Very hard, unpublished research-style maths problems
Better than 99% of 122 tested models. Run Aug 30, 2026.

Human-vote arenas

People compare two anonymous answers and pick the better one. Scores work like chess ratings.

ArenaRankScoreRange (95%)VotesUpdated
Text arena
Everyday chat, blind human votes
#75 of 41914411435–144715,038Oct 28, 2026
WebDev arena
Building web apps, blind human votes
#2 of 15217731764–17837,210Oct 28, 2026
Vision arena
Questions about images, blind human votes
#36 of 16712801270–12904,012Oct 19, 2026
Agent arena
Real multi-step agent sessions
#4 of 570.1370.117–0.15715,315Nov 2, 2026

Source: Arena (formerly LMArena) leaderboard dataset, CC BY 4.0. Fetched Nov 4, 2026.

Price

Input, per 1M tokens$10Output, per 1M tokens$50Blended (3 in : 1 out)$20Context window1.1MSpeedComing soon

List price from the OpenRouter models API (openai/gpt-6-astra), Nov 4, 2026. Providers can charge differently.