GPQA DiamondVerified
84.8%
GPQA Diamond (AA run)
Models / MiniMax
MiniMax M2.5
MiniMax · Released 2026-04-01
MiniMax's April 2026 agentic model. Superseded by M3 for open-weights work but still on leaderboards.
Prior MiniMax tier for coding agents when pinned to M2.5 endpoints.
Composite53
Context1M
Input / 1M$0.30
Output / 1M$1.2
Knowledge cutoff2026-02-01
Output speed68 t/s
Statussolid
Family profile
Best published score in each covered benchmark family.
64/100 avg
- Reasoning
- 85
- Agentic
- 35
- Long context
- 73
3 tested benchmark families
Benchmark placements
Per-benchmark positions
Where MiniMax M2.5 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 output speed | Performance | #88/ 195 | Artificial Analysis | 2026-10-12 | |
| GPQA Diamond | Reasoning | #99/ 184 | 85 | Artificial Analysis | 2026-09-28 |
| AA-LCR | Long context | #103/ 184 | 73 | Artificial Analysis | 2026-09-28 |
| Terminal-Bench | Agentic | #119/ 167 | 35 | Artificial Analysis | 2026-09-28 |
| Humanity's Last Exam | Reasoning | #127/ 186 | 21 | 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 M2.5 shows up, it's highlighted.
ReasoningMiniMax M2.5 · 85
- 196
- 295
- 395
Math
- 152
- 252
- 350
AgenticMiniMax M2.5 · 35
- 191
- 289
- 389
Long contextMiniMax M2.5 · 73
- 189
- 286
- 385
Newest receipts
5 recentLast 120 days
Benchmark rows added to the public ledger for MiniMax M2.5 in the last 120 days. Older rows live in the full table below.
| Benchmark | Family | Score | Source | Days ago |
|---|---|---|---|---|
| AA output speed | Performance | 68 t/s | Artificial Analysis | 10 |
| AA-LCR | Long context | 73.33% | Artificial Analysis | 24 |
| GPQA Diamond | Reasoning | 84.8% | Artificial Analysis | 24 |
| Humanity's Last Exam | Reasoning | 20.5% | Artificial Analysis | 24 |
| Terminal-Bench | Agentic | 34.85% | 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 M2.5 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 M2.5. Scores link to the original source; gaps mean no public row exists yet.
5 of 13 catalog benchmarks have a sourced row for MiniMax M2.5.
Reasoning
Humanity's Last ExamVerified
20.5%
HLE (AA run)
Agentic
Terminal-BenchVerified
34.85%
Terminal-Bench Hard (AA run)
Long context
AA-LCRVerified
73.33%
AA-LCR (AA run)
Performance
AA output speedVerified
68 t/s
Median output tokens/s (1k prompt, default provider)
Sources
Where the numbers come from
- Listed on the model atlas. Sourced scores that still have a public explainer are on the ledger.· checked 2026-06-09
- MiniMax· checked 2026-06-09
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 M2.5?
MiniMax M2.5 is a MiniMax model: MiniMax's April 2026 agentic model. Superseded by M3 for open-weights work but still on leaderboards.
What is MiniMax M2.5 for?
Prior MiniMax tier for coding agents when pinned to M2.5 endpoints.
What does MiniMax M2.5 cost?
MiniMax M2.5 lists $0.30 input and $1.2 output per 1M tokens. Pricing last checked 2026-06-09.
How large is the MiniMax M2.5 context window?
MiniMax M2.5 publishes a 1M token context window.
What is the VerdictPal status of MiniMax M2.5?
MiniMax M2.5 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.