MiniMax M2 vs Moondream 3 Preview
Updated October 21, 2026. No benchmark has tested both models yet. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
- Coding
- No data
- Overall
- 18.6
- Output $/M
- $1.02
- Coding
- No data
- Overall
- No data
- Output $/M
- $2.50
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
MiniMax M2 has the edge where output tokens cost 2.5x less, and where the context window is larger, 205K against 32K tokens.
No benchmark has tested both yet. MiniMax M2 has results on SWE-bench Verified, LMArena WebDev and LiveCodeBench (Artificial Analysis run); Moondream 3 Preview has none yet. Compare them on price and context instead.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
No call
Only MiniMax M2 has a AA Intelligence Index result, 18.6.
Directional only
Web design
Design Arena: websites
No call
Only MiniMax M2 has a Design Arena: websites result, 1143.
Directional only
Long documents
Listed context window
MiniMax M2
MiniMax M2 takes 205K tokens against 32K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
MiniMax M2
MiniMax M2 $0.0008, Moondream 3 Preview $0.0015 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only MiniMax M2 has the number, $0.702.
Directional only
At a glance
- Coding index
- MiniMax M2 not benchmarked, Moondream 3 Preview not benchmarked.
- Widest category
- None yet: no category has been tested on both models.
- Widest single gap
- None yet: no benchmark has tested both models.
- Output price
- MiniMax M2 $1.02 against Moondream 3 Preview $2.50 per million output tokens, 2.5x apart.
- Context window
- MiniMax M2 takes 205K tokens, Moondream 3 Preview 32K.
Benchmark by benchmark
Every board either model has a result on, a category to a fold; bars are each result on a 0 to 100 scale. Only one model is measured on each, so no row names a leader.
CodingDirectional onlySix of these boards feed the Vibe Coding Index8 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| SWE-bench Verified | 61.0% | Not measured | Only MiniMax M2 measured |
| LMArena WebDev | 1297 | Not measured | Only MiniMax M2 measured |
| LiveCodeBench (Artificial Analysis run) | 82.6% | Not measured | Only MiniMax M2 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 25.8% | Not measured | Only MiniMax M2 measured |
| Terminal-Bench | 30.0% | Not measured | Only MiniMax M2 measured |
| LMArena Text: coding | 1371 | Not measured | Only MiniMax M2 measured |
| LMArena Text: software and IT | 1373 | Not measured | Only MiniMax M2 measured |
| LMArena WebDev: HTML | 1306 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 30 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index1 board measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 18.6 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 1 board here | |||
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 86.8% | Not measured | Only MiniMax M2 measured |
| Vending-Bench 2 | $161 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 37 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index5 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 77.7% | Not measured | Only MiniMax M2 measured |
| Humanity's Last Exam (Artificial Analysis run) | 13.7% | Not measured | Only MiniMax M2 measured |
| LMArena Text: hard prompts | 1357 | Not measured | Only MiniMax M2 measured |
| LMArena Text: expert | 1338 | Not measured | Only MiniMax M2 measured |
| LMArena Text: hard prompts in English | 1376 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 16 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index4 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| AA Math Index | 78.3 | Not measured | Only MiniMax M2 measured |
| AIME 2025 (Artificial Analysis run) | 78.3% | Not measured | Only MiniMax M2 measured |
| LMArena Text: math | 1351 | Not measured | Only MiniMax M2 measured |
| LMArena Text: mathematical work | 1368 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 9 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index5 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 82.0% | Not measured | Only MiniMax M2 measured |
| LMArena Text: science | 1348 | Not measured | Only MiniMax M2 measured |
| LMArena Text: medicine and healthcare | 1367 | Not measured | Only MiniMax M2 measured |
| LMArena Text: legal and government | 1349 | Not measured | Only MiniMax M2 measured |
| LMArena Text: business and finance | 1336 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 12 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1143 | Not measured | Only MiniMax M2 measured |
| Design Arena: UI components | 1136 | Not measured | Only MiniMax M2 measured |
| Design Arena: game dev | 1118 | Not measured | Only MiniMax M2 measured |
| Design Arena: data viz | 1140 | Not measured | Only MiniMax M2 measured |
| Design Arena: 3D | 1098 | Not measured | Only MiniMax M2 measured |
| Design Arena: all code categories | 1135 | Not measured | Only MiniMax M2 measured |
| Design Arena: SVG | 1100 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 2 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index17 boards measured
| Benchmark | MiniMax M2 | Moondream 3 Preview | Leader |
|---|---|---|---|
| IFBench | 72.3% | Not measured | Only MiniMax M2 measured |
| Long context reasoning (AA-LCR) | 64.3% | Not measured | Only MiniMax M2 measured |
| LMArena Text | 1340 | Not measured | Only MiniMax M2 measured |
| LMArena Text: instruction following | 1329 | Not measured | Only MiniMax M2 measured |
| LMArena Text: creative writing | 1286 | Not measured | Only MiniMax M2 measured |
| LMArena Text: English | 1364 | Not measured | Only MiniMax M2 measured |
| LMArena Text: non-English | 1313 | Not measured | Only MiniMax M2 measured |
| LMArena Text: Chinese | 1366 | Not measured | Only MiniMax M2 measured |
| LMArena Text: German | 1357 | Not measured | Only MiniMax M2 measured |
| LMArena Text: French | 1336 | Not measured | Only MiniMax M2 measured |
| LMArena Text: Spanish | 1326 | Not measured | Only MiniMax M2 measured |
| LMArena Text: Russian | 1330 | Not measured | Only MiniMax M2 measured |
| LMArena Text: Polish | 1312 | Not measured | Only MiniMax M2 measured |
| LMArena Text: multi-turn | 1361 | Not measured | Only MiniMax M2 measured |
| LMArena Text: longer queries | 1331 | Not measured | Only MiniMax M2 measured |
| LMArena Text: writing and language | 1295 | Not measured | Only MiniMax M2 measured |
| LMArena Text: entertainment and media | 1298 | Not measured | Only MiniMax M2 measured |
| Neither model has a result on the other 29 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
MiniMax M2 costs less
Repository review
50K input + 3K output tokens
MiniMax M2 costs less
Agent loop
200K input + 10K output tokens
MiniMax M2 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | MiniMax M2 | Moondream 3 Preview |
|---|---|---|
| Maker | MiniMax | Moondream |
| Weights | Open weights | Open weights |
| Released | October 23, 2025 | October 15, 2026 |
| Context window | 205K | 32K |
| Input $/Mtok | $0.255 | $0.30 |
| Output $/Mtok | $1.02 | $2.50 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 1 | 0 |
| Cost per solved task | $0.702 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is MiniMax M2 better than Moondream 3 Preview for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. MiniMax M2 has the edge where output tokens cost 2.5x less, and where the context window is larger, 205K against 32K tokens.
Which is cheaper, MiniMax M2 or Moondream 3 Preview?
MiniMax M2 costs $0.255 per million input tokens and $1.02 per million output tokens; Moondream 3 Preview costs $0.30 and $2.50. Cost per solved task is known for MiniMax M2 only.
Which has the bigger context window, MiniMax M2 or Moondream 3 Preview?
MiniMax M2, with 205K tokens against 32K.
Has any benchmark tested both MiniMax M2 and Moondream 3 Preview?
No benchmark has tested both yet. MiniMax M2 has results on SWE-bench Verified, LMArena WebDev and LiveCodeBench (Artificial Analysis run); Moondream 3 Preview has none yet. Compare them on price and context instead. Data as of October 21, 2026.