Compare / MiniMax-M2.5 vs MiniMax-M3

MiniMax-M2.5 vs MiniMax-M3

Which is cheaper, which is more capable, and which does better at each kind of work, from 21 results on 4 sources that measured both models.

MiniMaxMiniMax-M2.5 Released 12 Feb 2026 · $0.29 in · $1.20 out
MiniMaxMiniMax-M3 Released 1 Jun 2026 · $0.30 in · $1.20 out
MiniMax-M2.5 wins 1MiniMax-M3 wins 20

Of the 21 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).

MiniMax-M2.5 or MiniMax-M3? The short answer

Pick MiniMax-M3: it scores higher (Intelligence Index 29.2 against 22.8) at about the same price.

  • They cost about the same: $0.52 and $0.52 per million tokens (three input to one output).
  • MiniMax-M3 scores higher on the Artificial Analysis Intelligence Index: 29.2 against 22.8.
  • MiniMax-M3 is newer: released 1 Jun 2026, against 12 Feb 2026 for MiniMax-M2.5.
  • On third-party benchmarks, on the results both models have: MiniMax-M3 does better for agents and tool use, reasoning and knowledge, writing and sticking to the facts.

Head to head

22.8
Intelligence Index
29.2
$0.52
Price per million tokens, blendedlower is better
$0.52
200,000
Context window, tokens
1,048,576

Every fact side by side

MiniMax-M2.5MiniMax-M3
Price per million tokens$0.29 in · $1.20 out$0.30 in · $1.20 out
Blended (3 in : 1 out)$0.52$0.52
Cheapest host, blended$0.44 (Venice)$0.41 (CoreWeave, FP4)
Cached input, per million–$0.060 cached input
Intelligence Index22.829.2
Spring Prompt overall––
Output speed, tokens a second–146
Context window200,000 tokens1,048,576 tokens
Released12 Feb 20261 Jun 2026 (newer)
WeightsOpenOpen
Developer based inChinaChina

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 135 of 159 79 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects not measured 39 of 51
Reasoning and knowledgeHard questions across science, maths and general knowledge 91 of 187 45 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 107 of 169 67 of 169
Sticking to the factsSummarising without inventing things, and factual answers 101 of 139 91 of 139

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureMiniMax-M2.5MiniMax-M3
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 150 · 84 of 387 measured 1,391 1,440
Overall · Text factualityArena rating, higher is better 141 · 89 of 163 measured 1,386 1,434
Business, management and finance · TextArena rating, higher is better 142 · 68 of 380 measured 1,400 1,448
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 92 · 59 of 249 measured 22.8 29.2
GPQA Diamond% of questions, higher is better 81 · 14 of 232 measured 84.8% 92.9%
Humanity's Last Exam% of questions, higher is better 116 · 51 of 250 measured 20.5% 39.0%
OpenHands Index reported by OpenHands Index · 3 results
Average% resolved, higher is better 22 · 13 of 34 measured 45.2%openhands 57.2%openhands
Front end (SWE-Bench Multimodal)% resolved, higher is better 27 · 11 of 34 measured 25.0%openhands 36.8%openhands
Greenfield (Commit0)% resolved, higher is better 25 · 14 of 34 measured 12.5%openhands 25.0%openhands
Vending-Bench 2 reported by Vending-Bench 2 · 1 result
Money after a yearUS dollars, higher is better 57 · 40 of 58 measured $-23.16 $2,158
Show all 21 results (11 more)
MeasureMiniMax-M2.5MiniMax-M3
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 4 results
Creative writing · TextArena rating, higher is better 147 · 84 of 385 measured 1,357 1,406
Expert prompts · TextArena rating, higher is better 130 · 80 of 338 measured 1,425 1,470
Instruction following · TextArena rating, higher is better 147 · 76 of 387 measured 1,381 1,434
Writing, literature and language · TextArena rating, higher is better 146 · 77 of 386 measured 1,369 1,422
Artificial Analysis reported by Artificial Analysis · 4 results
IFBench% of instructions, higher is better 43 · 3 of 195 measured 71.6% 82.9%
Long-context reasoning (AA-LCR)% of questions, higher is better 98 · 16 of 241 measured 73.3% 83.0%
Terminal-Bench Hard% of tasks, higher is better 49 · 29 of 190 measured 34.8% 42.4%
τ²-bench telecom% of tasks, higher is better 20 · 46 of 193 measured 95.3% 88.9%
OpenHands Index reported by OpenHands Index · 3 results
Information gathering (GAIA)% resolved, higher is better 24 · 15 of 34 measured 47.9%openhands 66.7%openhands
Issue resolution (SWE-Bench)% resolved, higher is better 23 · 9 of 34 measured 72.6%openhands 76.4%openhands
Testing (SWT-Bench)% resolved, higher is better 18 · 4 of 34 measured 68.1%openhands 81.1%openhands

Quick answers

Is MiniMax-M2.5 better than MiniMax-M3?

Pick MiniMax-M3: it scores higher (Intelligence Index 29.2 against 22.8) at about the same price.

Which is cheaper, MiniMax-M2.5 or MiniMax-M3?

They cost about the same: $0.52 and $0.52 per million tokens (three input to one output).

Which is better for agents and tool use, MiniMax-M2.5 or MiniMax-M3?

MiniMax-M3 ranks 79 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 135 for MiniMax-M2.5.

Which is better for reasoning and knowledge, MiniMax-M2.5 or MiniMax-M3?

MiniMax-M3 ranks 45 of 187 models on sale for hard questions across science, maths and general knowledge, against 91 for MiniMax-M2.5.

Which is better for writing, MiniMax-M2.5 or MiniMax-M3?

MiniMax-M3 ranks 67 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 107 for MiniMax-M2.5.

Which is better for sticking to the facts, MiniMax-M2.5 or MiniMax-M3?

MiniMax-M3 ranks 91 of 139 models on sale for summarising without inventing things, and factual answers, against 101 for MiniMax-M2.5.

Which has the bigger context window, MiniMax-M2.5 or MiniMax-M3?

MiniMax-M3: 1,048,576 tokens, against 200,000 for MiniMax-M2.5.

Run MiniMax-M2.5 and MiniMax-M3 on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.