Across 15 comparable benchmarks, MiniMax M2.5 has better reported results on 11 and worse on 2, with 2 tied. 1 further shared result is listed below, each with the reason it is not counted.
Strong evidence15 of 16 shared results countedHow we compare
Claude Sonnet 4 is no longer available through its official API. API prices for this version are not shown.
Price basis: Listed rates vs listed rates. The price section below shows the applicable periods.
| Fact | MiniMax M2.5 | Claude Sonnet 4 |
|---|---|---|
| Context window | 128K tokens | 200K tokens |
| Input / 1M | $0.300 | — |
| Output / 1M | $1.20 | — |
| Blended (4:1) | $0.480 | — |
| Status | Available | Retired |
| Released | February 12, 2026 | May 23, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 10 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
MiniMax M2.5 ahead on 2 of 3 counted
MiniMax M2.5 ahead on 1 of 1 counted
MiniMax M2.5 ahead on 1 of 1 counted
MiniMax M2.5 ahead on 1 of 2 counted
MiniMax M2.5 ahead on 2 of 2 counted
MiniMax M2.5 ahead on 2 of 3 counted
No shared comparable results for Research & search, Long context, Multimodal, Cybersecurity.
Published scores, with matching modes where available. Gaps show MiniMax M2.5 relative to Claude Sonnet 4. Reading the results
| Benchmark | MiniMax M2.5 | Claude Sonnet 4 | Gap |
|---|---|---|---|
| SWE-Bench Pro - PublicRepository Engineering · Accuracy | 55.4Thinking · tools | 42.7≠Normal · tools | +12.7 pts#32 vs #58 of 66 |
| SWE-bench VerifiedRepository Engineering · Accuracy | 80.2Thinking · tools | 80.2≠Parallel Thinking · tools | =#15 vs #15 of 108 |
| Terminal Bench HardAgentic Development · Accuracy | 34.8Thinking · tools | 31.1Thinking · tools | +3.7 pts#50 vs #66 of 157 |
| ARC-AGI-1Abstract Generalization · Score (%) | 63.7Thinking · no tools | 40.0Thinking · no tools | +23.7 pts#56 vs #69 of 87 |
| ARC-AGI-2Abstract Generalization · Score (% solved); cost per task (USD)Not counted · Speed and cost telemetry is reported, not scored | 4.9Thinking · no tools | 5.9Thinking · no tools | 0.82×#66 vs #64 of 80 |
| AIME2025Mathematics · Accuracy | 86.3Thinking · no tools | 85.0≠Deep Thinking · tools | +1.3 pts#46 vs #48 of 86 |
| CritPtScientific Reasoning · Score | 1.1Thinking · no tools | 1.1≠Normal · no tools | =#103 vs #103 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 85.2Thinking · no tools | 83.8≠Deep Thinking · tools | +1.4 pts#80 vs #91 of 200 |
| Claw BenchTool Orchestration · Accuracy | 92.1Thinking · tools | 77.8Thinking · tools | +14.3 pts#4 vs #23 of 29 |
| Pinch BenchTool Orchestration · Accuracy | 87.8Thinking · tools | 80.5Thinking · tools | +7.3 pts#7 vs #23 of 38 |
| Creative WritingWriting | 1362Normal · no tools | 1483Normal · no tools | 0.92×#80 vs #62 of 121 |
| IF BenchInstruction Following · Accuracy | 70.0Thinking · tools | 55.0Thinking · tools | +15.0 pts#25 vs #34 of 38 |
| GDPval-AACross-industry Work · Accuracy | 36.0Thinking · no tools | 33.0Thinking · no tools | +3.0 pts#10 vs #12 of 14 |
| HLEKnowledge Exams · Accuracy | 19.4Thinking · no tools | 9.6Thinking · no tools | +9.8 pts#106 vs #130 of 151 |
| ECICapability Indices · ECI score (capability index, higher is better) | 147Reported best | 142Reported best | 1.04×#65 vs #91 of 168 |
| LiveBenchCross-capability Suites · Accuracy | 60.1Deep Thinking · no tools | 61.3≠Thinking · no tools · 64K | −1.1 pts#85 vs #83 of 106 |
Repository Engineering · Accuracy
Repository Engineering · Accuracy
Agentic Development · Accuracy
Mathematics · Accuracy
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Tool Orchestration · Accuracy
Tool Orchestration · Accuracy
Writing
Instruction Following · Accuracy
Cross-industry Work · Accuracy
A dash means no result is available.
| Benchmark | MiniMax M2.5 | Claude Sonnet 4 |
|---|---|---|
| AIME 2024 | — | 43.4 |
| FrontierMath | — | 4.1 |
| FrontierMath - Tier 4 | — | 0.0 |
| IMO-ProofBench | — | 27.1 |
| IMO-ProofBench Advanced | — | 4.8 |
| MATH-500 | — | 99.1 |
| MMLU-Pro | — | 84.0 |
| Aider-Polyglot | — | 61.3 |
| CodeClash | — | 1223 |
| IOI 2024 (Vals v1) | — | 9.7 |
| IOI 2025 (Vals v1) | — | 4 |
| LiveCodeBench | — | 48.5 |
| SWE-bench Multilingual | 68.3 | — |
| SWE-Bench Pro - Commercial | — | 9.1 |
| Terminal Bench 2.0 | 51.7 | — |
| Terminal-Bench 1.0 | — | 41.3 |
| Vibe Code Bench v1.1 | 14.8 | — |
| OSWorld-Verified | — | 42.2 |
| PinchBench v2 | — | 48.8 |
| Fiction.liveBench | — | 46.9 |
| MMMU | — | 74.4 |
| MMMU-Pro | — | 62.4 |
| VPCT | — | 34.0 |
| SAGE | — | 35.0 |
| τ²-Bench | — | 52.0 |
| τ²-Bench - Telecom | 97.8 | — |
| τ³-Banking | — | 16.7 |
| SimpleBench | — | 45.5 |
| Vending-Bench 2 | -23.2 | — |
| BrowseComp | 76.3 | — |
| MedCode | — | 35.0 |
| MedScribe | — | 72.4 |
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
A blended cost needs a published input and output price on both sides, and one of these two does not have them yet.
MiniMax M2.5: Listed text rates; pricing tier not confirmed.
Claude Sonnet 4: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | MiniMax M2.5 | Claude Sonnet 4 | Ratio |
|---|---|---|---|
| Input | $0.30 | — | — |
| Output | $1.20 | — | — |
| Cached input | — | — | — |
| Cache write | — | — | — |
| Blended | $0.48 | — | — |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Lower prices are highlighted. The cache hit rate applies to input tokens only; models without a published cached-input rate use their normal input price for those tokens.Effective dates: MiniMax M2.5 2026-11-14; Claude Sonnet 4 2026-11-14.Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
MiniMax M2.5 relative to Claude Sonnet 4
Summary: 2 of 2 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 2 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
Effort, protocol, aggregation and source match; other run conditions may differ.
MiniMax M2.5
Claude Sonnet 4
Effort, protocol, aggregation and source match; other run conditions may differ.
MiniMax M2.5
Claude Sonnet 4
| Benchmark | MiniMax M2.5 | Claude Sonnet 4 |
|---|---|---|
| Vibe Code Bench v1.1 | Not reported |
| Attribute | MiniMax M2.5 | Claude Sonnet 4 |
|---|---|---|
| Availability | Available | Retired |
| Provider | MiniMaxAI | Anthropic |
| Model type | Reasoning model | Reasoning model |
| First release | February 12, 2026 | May 23, 2025 |
| Context window | 128K tokens | 200K tokens |
| Max output | — | 64K tokens |
| Total parameters | 229B | — |
| Active parameters | 10B | — |
| Architecture | Mixture of Experts (MoE) | Undisclosed |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual |
| Input modalities | Text | Text, Image |
| Output modalities | Text | Text |
| Weights | Open weights | Closed / API only |
| Licence | MINIMAX MODEL LICENSE | — |
| Commercial use | Commercial use allowed | Proprietary |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used MiniMax M2.5 or Claude Sonnet 4? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.