GPT-6 Astra vs MiniMax-M3

At a glance

• Capability index

  • GPT-6 Astra166.4
  • MiniMax-M3146.9

  GPT-6 Astra is 19.5 points higher.

• Price per 1M tokens

  • GPT-6 Astra$20.00
  • MiniMax-M3$0.525

  MiniMax-M3 is 38 times cheaper.

• Context window

  • GPT-6 Astra1.1M tokens
  • MiniMax-M31M tokens

  GPT-6 Astra takes in 5% more.

• Released

  • GPT-6 AstraSep 4, 2026
  • MiniMax-M3Jun 1, 2026

  GPT-6 Astra is 3 months newer.

Where each scores higher

On the 18 tests both have taken, each on its own scale. A gap of a point or two can sit within a test’s margin of error.

4 more tests have results for only one of them; the table below lists every result.

Everything side by side

Comparison of GPT-6 Astra vs MiniMax-M3

Measure	GPT-6 AstraOpenAI	MiniMax-M3MiniMax
Model
Maker	OpenAI	MiniMax
Sold here by	OpenAI	MiniMax (minimax.io)
API model ID	gpt-6-astra	MiniMax-M3
Released	Sep 4, 2026	Jun 1, 2026
Knowledge cutoff	Apr 30, 2026	Jan 2025
Weights	Closed: hosted access only	Open: downloadable
Developer’s country	United States	China
Hosts selling it	29including OpenAI directly	56including MiniMax directly
Training compute	1.0 × 10²⁷ FLOPEpoch AI estimate	Not published
Test results
Capability index	166.4 (best of these models)#2 of 274	146.9#68 of 274
LiveBench overall	82.2% (best of these models)#4 of 66	67.3%#61 of 66
Arena text	1477 (best of these models)#29 of 413	1440#96 of 413
LiveBench coding	80.4% (best of these models)#20 of 66	68.2%#64 of 66
Arena WebDev	1788 (best of these models)#2 of 138	1482#58 of 138
Arena coding	1543 (best of these models)#7 of 408	1494#86 of 408
LiveBench mathematics	96.8% (best of these models)#4 of 66	77.0%#65 of 66
Mock AIME 2024–2025	100.0% (best of these models)#1 of 297	71.1%#133 of 297
FrontierMath Tiers 1–3	93.7%#1 of 114	Not tested
FrontierMath Tier 4	97.6%#2 of 70	Not tested
Arena math	1490 (best of these models)#27 of 396	1432#102 of 396
SimpleQA Verified	75.6%#1 of 86	Not tested
LiveBench agentic coding	57.3% (best of these models)#21 of 66	40.7%#60 of 66
DeepSWE	74.1%#1 of 69	Not tested
LiveBench language	89.4% (best of these models)#4 of 66	76.8%#45 of 66
LiveBench instructions	75.6% (best of these models)#8 of 66	57.5%#61 of 66
Arena creative writing	1449 (best of these models)#39 of 411	1406#99 of 411
Arena instruction following	1474 (best of these models)#30 of 413	1434#88 of 413
LiveBench data analysis	83.0% (best of these models)#1 of 66	76.2%#34 of 66
LiveBench reasoning	92.7% (best of these models)#1 of 66	74.5%#61 of 66
GPQA Diamond	95.8% (best of these models)#1 of 319	90.9%#34 of 319
Chess puzzles	72.0% (best of these models)#1 of 227	14.0%#113 of 227
Arena hard prompts	1502 (best of these models)#30 of 413	1463#90 of 413
Price per million tokens
Input	$10.00	$0.30 (best of these models)
Output	$50.00	$1.20 (best of these models)
Blended, 3 input to 1 output	$20.00	$0.525 (best of these models)
Cached input	$1.00	$0.06 (best of these models)
Writing to the cache	$12.50	Not reported
Long requests	$20.00 in, $75.00 outabove 272K tokens	$0.60 in, $2.40 outabove 512K tokens
The maker’s own price	This listing	This listing
Limits
Context window	1.1M tokens (best of these models)	1M tokens
Max input	922K tokens (best of these models)	512K tokens
Max output	128K tokens	512K tokens (best of these models)
Features
Reads	images, PDFs and text	images, text and video
Produces	text	text
Reasoning	Yeseffort low, medium, high, xhigh, max	Yeseffort low, medium, high; can be switched off; thinking budget can be set
Tool calling	Yes	Yes
Structured output	Yes, such as JSON	Yes, such as JSON
File attachments	Yes	Yes
Temperature setting	Fixed by the provider	Supported

Bold marks the better value in each row. Prices are each model’s own list price, or the middle price across hosts when the maker does not sell it, blended as three input tokens for every output token. Test results as published by LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Datacurve DeepSWE, compiled by Epoch AI (CC BY 4.0), Arena (CC BY 4.0); model details and prices from models.dev (MIT). Confirm pricing with the provider before use.