GPQA DiamondVerified
92.9%
GPQA Diamond (AA run)
Models / MiniMax
MiniMax M3
MiniMax · Released 2026-05-31
MiniMax's May 2026 open-weights release. Terminal-Bench 2.0 66.0% — the #10 score globally and the strongest non-frontier-tier result on the public ledger. 1M context. Pay-as-you-go list $0.30 / $1.20 per 1M tokens on inputs up to 512K.
Default open-weights model for self-hosted agent infrastructure. Cheaper than DeepSeek V4 Pro Max on output ($2 vs $1.60 — well, slightly higher), with comparable Terminal-Bench scores.
Composite65
Context1M
Input / 1M$0.30
Output / 1M$1.2
Knowledge cutoff2026-03-01
Output speed107 t/s
Statussolid
Family profile
Best published score in each covered benchmark family.
74/100 avg
- Reasoning
- 93
- Coding
- 47
- Long context
- 83
3 tested benchmark families
Editor's note
MiniMax M3 is the dark horse of the open-weights tier — Terminal-Bench 2.0 66.0% puts it ahead of several proprietary models. The 1M context is the real differentiator for long-document work.Benchmark placements
Per-benchmark positions
Where MiniMax M3 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.
| Benchmark | Family | Rank | Score | Source | Date |
|---|---|---|---|---|---|
| AA-LCR | Long context | #15/ 184 | 83 | Artificial Analysis | 2026-09-28 |
| GPQA Diamond | Reasoning | #17/ 184 | 93 | Artificial Analysis | 2026-09-28 |
| Humanity's Last Exam | Reasoning | #48/ 186 | 39 | Artificial Analysis | 2026-09-28 |
| AA output speed | Performance | #54/ 195 | Artificial Analysis | 2026-10-12 | |
| SciCode | Coding | #58/ 97 | 47 | Artificial Analysis | 2026-09-28 |
Family context
Strongest rows in each family
The three highest-scoring models with pages in each capability family. Where MiniMax M3 shows up, it's highlighted.
ReasoningMiniMax M3 · 93
- 196
- 295
- 395
Math
- 152
- 252
- 350
CodingMiniMax M3 · 47
- 174
- 273
- 370
Long contextMiniMax M3 · 83
- 189
- 286
- 385
Newest receipts
5 recentLast 120 days
Benchmark rows added to the public ledger for MiniMax M3 in the last 120 days. Older rows live in the full table below.
| Benchmark | Family | Score | Source | Days ago |
|---|---|---|---|---|
| AA output speed | Performance | 107 t/s | Artificial Analysis | 10 |
| AA-LCR | Long context | 83% | Artificial Analysis | 24 |
| GPQA Diamond | Reasoning | 92.9% | Artificial Analysis | 24 |
| Humanity's Last Exam | Reasoning | 39% | Artificial Analysis | 24 |
| SciCode | Coding | 47.1% | Artificial Analysis | 24 |
Family coverage
Best score per family
A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.
Price vs. performance
Where MiniMax M3 sits on the scatter
Hover any dot for name, score, and input price. Keyboard: tab through the top twelve, or use the ranking below.
AnthropicOpenAIGoogleAlibabaMetaZhipuxAIDeepSeekMiniMaxXiaomiMoonshotOtherMeituanMistralNVIDIACohere
Full benchmark ledger
Every catalog benchmark for MiniMax M3. Scores link to the original source; gaps mean no public row exists yet.
5 of 13 catalog benchmarks have a sourced row for MiniMax M3.
Reasoning
Humanity's Last ExamVerified
39%
HLE (AA run)
Coding
SciCodeVerified
47.1%
SciCode (AA run)
Long context
AA-LCRVerified
83%
AA-LCR (AA run)
Performance
AA output speedVerified
107 t/s
Median output tokens/s (1k prompt, default provider)
Changelog
Version history
- Pay-as-you-go list re-checked. Image and video input are on the launch post.
- Released MiniMax M3 — Terminal-Bench 2.0 66.0%, 1M context.
- MiniMax M2.7 retired as the flagship open-weights model.
Sources
Where the numbers come from
- MiniMax M3 model card· checked 2026-06-01
- MiniMax API pricing· checked 2026-09-28
- MiniMax M3 launch· checked 2026-09-28
- Terminal-Bench 2.0 leaderboard· checked 2026-06-01
Questions this model page answers
Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.
What is MiniMax M3?
MiniMax M3 is a MiniMax model: MiniMax's May 2026 open-weights release. Terminal-Bench 2.0 66.0% — the #10 score globally and the strongest non-frontier-tier result on the public ledger. 1M context. Pay-as-you-go list $0.30 / $1.20 per 1M tokens on inputs up to 512K.
What is MiniMax M3 for?
Default open-weights model for self-hosted agent infrastructure. Cheaper than DeepSeek V4 Pro Max on output ($2 vs $1.60 — well, slightly higher), with comparable Terminal-Bench scores.
What does MiniMax M3 cost?
MiniMax M3 lists $0.30 input and $1.2 output per 1M tokens. Pricing last checked 2026-09-28.
How large is the MiniMax M3 context window?
MiniMax M3 publishes a 1M token context window.
What is the VerdictPal status of MiniMax M3?
MiniMax M3 carries Solid status on the model atlas.
The Pack · Editorial newsletter
Know when the rules move.
One short email when a university changes its AI rule for theses, or a research tool changes its pricing or privacy terms. No tracking, no third-party analytics. Unsubscribe in one click.