Model comparison
MiniMax M2.7 vs MiniMax M3
Head-to-head evidence from 21 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Verified leaderboard positions: MiniMax M2.7 unranked; MiniMax M3 #18
BenchAlign evidence: MiniMax M2.7 supported; MiniMax M3 supported. Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. MiniMax M2.7 and MiniMax M3 share 21 comparable benchmark results. 2 of 8 categories are comparable. 16 results are unique to MiniMax M2.7; 24 to MiniMax M3.
Updated July 16, 2026- Shared results
- 21
- MiniMax M2.7 only
- 16
- MiniMax M3 only
- 24
- Comparable categories
- 2 / 8
Pick MiniMax M3 if you want the stronger benchmark profile. MiniMax M2.7 only becomes the better choice if its workflow or ecosystem matters more than the raw scoreboard.
Confidence note. This is a partial-evidence comparison with 21 shared benchmark results across 6 evidence categories; 2 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
MiniMax M3 is clearly ahead on the provisional aggregate, 70 to 52. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
MiniMax M3's sharpest advantage is in coding, where it averages 72.2 against 53.3. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 57% to 66%.
MiniMax M3 gives you the larger context window at 1M, compared with 200K for MiniMax M2.7.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | MiniMax M2.7 | Δ | MiniMax M3 |
|---|---|---|---|
| Coding | MiniMax M2.753.3 | Margin→ 18.9 | MiniMax M372.2 |
| Agentic | MiniMax M2.757.0 | Margin→ 15.3 | MiniMax M372.3 |
| Math | MiniMax M2.7Not measured | MarginNo overlap | MiniMax M385.7 |
| Multimodal | MiniMax M2.7Not measured | MarginNo overlap | MiniMax M364.9 |
Decisive benchmark drivers
The largest measured benchmark gaps in this matchup, with exact reported values.
More
- Source ↗
Terminal-Bench 2.0
AgenticA 57%B 66%Winner: MiniMax M3Δ 9Terminal-Bench 2.0: MiniMax M2.7 scored 57%; MiniMax M3 scored 66%. MiniMax M3 wins this benchmark. - Source ↗
SWE-bench Pro
CodingA 56.2%B 59%Winner: MiniMax M3Δ 2.8SWE-bench Pro: MiniMax M2.7 scored 56.2%; MiniMax M3 scored 59%. MiniMax M3 wins this benchmark.
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | MiniMax M2.7 | MiniMax M3 | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | MiniMax M2.7$0.3 input / $1.2 output | MiniMax M3$0.3 input / $1.2 output | Listed prices are equal. |
| Generation speedtokens per second | MiniMax M2.745 tok/s | MiniMax M3Not available | A complete speed comparison is not available. |
| First-answer latencyseconds to first token | MiniMax M2.72.53 s | MiniMax M3Not available | A complete latency comparison is not available. |
| Context windowmaximum listed tokens | MiniMax M2.7200K | MiniMax M31M | MiniMax M3 lists the larger context window. |
Benchmark Deep Dive
AgenticMiniMax M3 wins21 benchmarks
| Benchmark | MiniMax M2.7 | MiniMax M3 | Result |
|---|---|---|---|
| Terminal-Bench 2.0Source | 57% | 66% | MiniMax M3 leads |
| τ²-bench resultsSource | 84.8% | 88.9% | MiniMax M3 leads |
| ToolathlonSource | 46.3% | — | Not comparable |
| MLE-Bench LiteSource | 66.6% | — | Not comparable |
| MM-ClawBenchSource | 62.7% | — | Not comparable |
| Claw-EvalSource | 48.7% | 74.5% | MiniMax M3 leads |
| AA Agentic IndexSource | 25.6% | 35.4% | MiniMax M3 leads |
| APEX-Agents-AASource | 10.6% | — | Not comparable |
| GDPval-AASource | 32.9% | 44.7% | MiniMax M3 leads |
| GDPval-AASource | 1158 | 1395 | MiniMax M3 leads |
| Gert LabsSource | 40.40% | — | Not comparable |
| BrowseCompSource | — | 83.5% | Not comparable |
| OSWorld-VerifiedSource | — | 70.1% | Not comparable |
| MCP AtlasSource | — | 74.2% | Not comparable |
| GDPval rubricsSource | — | 74.7% | Not comparable |
| BankerToolBenchSource | — | 76.1% | Not comparable |
| ResearchClawBenchSource | — | 19.8% | Not comparable |
| OSWorld 2.0Source | — | 4.6% | Not comparable |
| AA BriefcaseSource | — | 1110 | Not comparable |
| AA EnterpriseOps-GymSource | — | 32.1% | Not comparable |
| AA Harvey LABSource | — | 6.7% | Not comparable |
CodingMiniMax M3 wins18 benchmarks
| Benchmark | MiniMax M2.7 | MiniMax M3 | Result |
|---|---|---|---|
| SWE-bench Verified*Source | 75.4% | — | Not comparable |
| SWE-bench ProSource | 56.2% | 59% | MiniMax M3 leads |
| SWE-RebenchSource | 51.9% | — | Not comparable |
| SWE MultilingualSource | 76.5% | — | Not comparable |
| Multi-SWE BenchSource | 52.7% | — | Not comparable |
| VIBE-ProSource | 55.6% | — | Not comparable |
| NL2RepoSource | 39.8% | 42.1% | MiniMax M3 leads |
| Vibe Code BenchSource | 27.04% | — | Not comparable |
| React Native EvalsSource | 71.4% | — | Not comparable |
| AA Coding IndexSource | 52.6% | 58.6% | MiniMax M3 leads |
| Terminal-Bench HardSource | 39.4% | 42.4% | MiniMax M3 leads |
| AA-SciCodeSource | 47.0% | 45.4% | MiniMax M2.7 leads |
| SWE-bench VerifiedSource | — | 80.5% | Not comparable |
| Terminal-Bench 2.0Source | — | 66.0% | Not comparable |
| VIBE V2Source | — | 50.1% | Not comparable |
| SVG-BenchSource | — | 63.7% | Not comparable |
| KernelBench HardSource | — | 28.8% | Not comparable |
| AA Terminal-Bench 2.1Source | — | 65.2% | Not comparable |
Reasoning2 benchmarks
Knowledge9 benchmarks
| Benchmark | MiniMax M2.7 | MiniMax M3 | Result |
|---|---|---|---|
| GPQA-DSource | 87.0% | — | Not comparable |
| MMLU-Pro (Arcee)Source | 80.8% | — | Not comparable |
| Artificial Analysis Intelligence IndexSource | 38.1% | 44.4% | MiniMax M3 leads |
| AA-GPQA DiamondSource | 87.4% | 92.9% | MiniMax M3 leads |
| AA-HLESource | 28.1% | 37.1% | MiniMax M3 leads |
| AA-Omniscience IndexSource | 0.7% | 1.4% | MiniMax M3 leads |
| AA-Omniscience AccuracySource | 26.1% | 15.0% | MiniMax M2.7 leads |
| AA-Omniscience Hallucination RateSource | 34.4% | 16.1% | MiniMax M3 leads |
| AA Openness IndexSource | — | 33.3% | Not comparable |
Math2 benchmarks
Multimodal8 benchmarks
| Benchmark | MiniMax M2.7 | MiniMax M3 | Result |
|---|---|---|---|
| GDPval-AASource | 1495 | — | Not comparable |
| Design Arena WebsiteSource | 1279 | 1294 | MiniMax M3 leads |
| OfficeQA ProSource | — | 45.1% | Not comparable |
| OmniDocBench 1.5Source | — | 91.6% | Not comparable |
| MMMU-ProSource | — | 78.1% | Not comparable |
| VideoMMMUSource | — | 84.6% | Not comparable |
| Video-MME (with subtitle)Source | — | 85.4% | Not comparable |
| AA-MMMU-ProSource | — | 78.6% | Not comparable |
Inst. Following1 benchmarks
| Benchmark | MiniMax M2.7 | MiniMax M3 | Result |
|---|---|---|---|
| AA-IFBenchSource | 75.7% | 82.9% | MiniMax M3 leads |
Frequently Asked Questions (3)
Which is better, MiniMax M2.7 or MiniMax M3?
MiniMax M3 is ahead on BenchLM's provisional leaderboard, 70 to 52. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 57% and 66%.
Which is better for coding, MiniMax M2.7 or MiniMax M3?
MiniMax M3 has the edge for coding in this comparison, averaging 72.2 versus 53.3. Inside this category, AA Coding Index is the benchmark that creates the most daylight between them.
Which is better for agentic tasks, MiniMax M2.7 or MiniMax M3?
MiniMax M3 has the edge for agentic tasks in this comparison, averaging 72.3 versus 57. Inside this category, GDPval-AA is the benchmark that creates the most daylight between them.
Related Comparisons
Explore More
The AI models change fast. We track them for you.
A weekly brief for engineers and researchers covering new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.