PodcastIntelligence Snacks · Weekly conversations about AI, software and business, from the team behind Normie Mode.Listen →
Models · OpenAI

GPT-6 Astra

OpenAI's GPT model, released 3 Sept 2026.

Top three for

7 tasks, including research, coding and building apps, excel and spreadsheets.

Evidence

Independently tested. 27 independent results so far.

Cost

$55 per 1,000 typical requests, if you use it directly.

Reads at once

1,575 pages of about 500 words.

Where you can use it

The apps and plans that include it, cheapest first.

App and planPriceAccess
ChatGPT Plus$20/monthIncluded
Only in ChatGPT's Work mode and Codex, not normal chat
ChatGPT Pro$100/month
From $100 a month; a $200 tier has higher limits
Included, the default
Called GPT-6 Pro in the app

Where it ranks, task by task

Its place on each task's score, among the models we compare that have enough results to score.

TaskPositionTask score
Best AI for research1st of 1180
Best AI for coding and building apps2nd of 1079
Best AI for Excel and spreadsheets2nd of 894
Best AI for business3rd of 865
Best AI for legal questions3rd of 871
Best AI for accounting and finance3rd of 865
Best AI for presentations and PowerPoint3rd of 853
Best AI for maths and science4th of 1278
Best AI for studying and homework7th of 1273
Best AI for documents, PDFs and notes7th of 1364
Best AI for planning and personal admin8th of 1443
Best AI for translation and languages9th of 1440
Best AI for CVs, job applications and interviews9th of 1435
Best AI for writing10th of 1434
Best AI for health questions12th of 1432

Test results

Every test we use that has a result for this exact version.

TestResultPosition
Research-level maths
Epoch AI
94%1st of 11
Graduate-level science questions
NYU and others
96%1st of 11
Answering short factual questions correctly
Google DeepMind
76%1st of 11
Would a real maintainer accept its code?
Cognition
53%1st of 9
Building features in real software projects
Datacurve
74%1st of 8
Pulling information out of documents into tables
DTBench authors
97%2nd of 13
Professional tasks in banking, consulting and law
Mercor
65%2nd of 8
Humanity's Last Exam
Center for AI Safety, Scale AI
55%1st of 4
Competition maths
Epoch AI
100%3rd of 12
Real paid freelance projects
Scale AI, Center for AI Safety
21%1st of 2
People's votes on coding questions
LMArena
rating 14888th of 14
People's votes on back-and-forth conversations
LMArena
rating 14578th of 14
People's votes on business and finance questions
LMArena
rating 14459th of 14
People's votes on expert questions
LMArena
rating 14869th of 14
People's votes on following instructions
LMArena
rating 14549th of 14
People's votes on legal questions
LMArena
rating 14609th of 14
People's votes on questions in other languages
LMArena
rating 14319th of 14
People's votes on maths questions
LMArena
rating 14639th of 13
People's votes on creative writing
LMArena
rating 142310th of 14
People's votes on long, detailed requests
LMArena
rating 146010th of 14
People's votes overall
LMArena
rating 144310th of 14
People's votes on science questions
LMArena
rating 145710th of 14
People's votes on writing and language
LMArena
rating 144210th of 14
People's votes on health questions
LMArena
rating 142311th of 13

Details

Maker
OpenAI
Released
3 Sept 2026
Status
Current
Name checked
Confirmed with OpenAI
Intelligence Snacks newsletter

The big AI ideas each week, in your inbox