Across 14 comparable benchmarks, Kimi K2.5 has better reported results on 12 and worse on 2.
Strong evidence14 of 14 shared results countedHow we compare
Kimi K2.5 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 | M2.1 | Kimi K2.5 |
|---|---|---|
| Context window | 200K tokens | 256K tokens |
| Input / 1M | $0.300 | — |
| Output / 1M | $1.20 | — |
| Blended (4:1) | $0.480 | — |
| Status | Available | Retired |
| Released | December 23, 2025 | January 27, 2026 |
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
Kimi K2.5 ahead on 4 of 4 counted
Kimi K2.5 ahead on 1 of 1 counted
No fixed-scale average — see the rows in Benchmark results
Kimi K2.5 ahead on 1 of 1 counted
Kimi K2.5 ahead on 3 of 4 counted
Kimi K2.5 ahead on 1 of 1 counted
Kimi K2.5 ahead on 1 of 1 counted
No shared comparable results for Long context, Multimodal, Professional work & productivity, Cybersecurity.
Published scores, with matching modes where available. Gaps show M2.1 relative to Kimi K2.5. Reading the results
| Benchmark | M2.1 | Kimi K2.5 | Gap |
|---|---|---|---|
| SWE-Bench Pro - PublicRepository Engineering · Accuracy | 32.6Thinking · tools | 50.7Thinking · tools | −18.1 pts#64 vs #49 of 66 |
| SWE-bench VerifiedRepository Engineering · Accuracy | 74.8Thinking · no tools | 76.8≠Thinking · tools | −2.0 pts#38 vs #28 of 108 |
| Terminal Bench 2.0Agentic Development · Accuracy | 47.9Thinking · tools | 50.8Thinking · tools | −2.9 pts#34 vs #33 of 46 |
| Terminal Bench HardAgentic Development · Accuracy | 28.8Thinking · tools | 34.8Thinking · tools | −6.0 pts#72 vs #50 of 157 |
| SimpleBenchCommonsense · Score (AVG@5) | 34.7Normal · no tools | 46.8≠Thinking · no tools | 0.74×#84 vs #68 of 106 |
| AIME2025Mathematics · Accuracy | 81.0Thinking · no tools | 96.1Thinking · no tools | −15.1 pts#55 vs #19 of 86 |
| CritPtScientific Reasoning · Score | 0.3Thinking · no tools | 3.1Thinking · no tools | 0.10×#124 vs #78 of 134 |
| GPQA DiamondScientific Reasoning · Accuracy | 81.0Thinking · no tools | 87.6Thinking · no tools | −6.6 pts#106 vs #67 of 200 |
| MedCodeClinical Workflows · Accuracy (%) | 34.1Unspecified Reasoning · tools | 39.3≠Thinking · tools | −5.2 pts#76 vs #64 of 97 |
| MedScribeClinical Workflows · Rubric pass rate (%) | 80.8Unspecified Reasoning · tools | 76.4≠Thinking · tools | +4.3 pts#51 vs #69 of 99 |
| Pinch BenchTool Orchestration · Accuracy | 84.3Thinking · tools | 84.8Thinking · tools | −0.5 pts#19 vs #18 of 38 |
| BrowseCompFact Finding · Accuracy | 47.4Thinking · tools | 60.6≠Thinking · tools + web | −13.2 pts#49 vs #42 of 61 |
| HLEKnowledge Exams · Accuracy | 22.0Thinking · no tools | 50.2≠Thinking · tools | −28.2 pts#95 vs #48 of 151 |
| MMLU-ProKnowledge Exams · Accuracy | 88.0Thinking · no tools | 78.5Thinking · no tools | +9.5 pts#7 vs #68 of 153 |
Repository Engineering · Accuracy
Repository Engineering · Accuracy
Agentic Development · Accuracy
Agentic Development · Accuracy
Commonsense · Score (AVG@5)
Mathematics · Accuracy
Scientific Reasoning · Score
Scientific Reasoning · Accuracy
Clinical Workflows · Accuracy (%)
Clinical Workflows · Rubric pass rate (%)
Tool Orchestration · Accuracy
Fact Finding · Accuracy
A dash means no result is available.
| Benchmark | M2.1 | Kimi K2.5 |
|---|---|---|
| AIME 2026 | — | 95.8 |
| FrontierMath - Tier 4 | — | 4.2 |
| IMO-AnswerBench | — | 81.8 |
| Riemann-bench | — | 12.0 |
| APEX-SWE | — | 30.3 |
| Code Migration | — | 7.0 |
| FrontierSWE v1 | — | 23.3 |
| LiveCodeBench | — | 85.0 |
| SWE-bench Multilingual | — | 73.0 |
| Vibe Code Bench v1.1 | — | 17.5 |
| WeirdML v2 | — | 45.6 |
| Claw Bench | — | 81.7 |
| MCP-Atlas | — | 64.4 |
| OSWorld-Verified | — | 63.3 |
| PinchBench v2 | — | 54.6 |
| Toolathlon-Verified | — | 33.0 |
| ECI | — | 148 |
| LiveBench | — | 69.1 |
| Text Arena (Coding) | — | 1431 |
| Vals Index | — | 26.3 |
| Creative Writing | — | 1579 |
| IF Bench | 70.0 | — |
| Context Arena | — | 59.2 |
| Fiction.liveBench | — | 86.1 |
| LongBench v2 | — | 61.0 |
| MMMU-Pro | — | 75.4 |
| EMB (Excel Modeling) | — | 28.5 |
| Finance Agent v2 | — | 35.8 |
| GDPval-AA | — | 40.0 |
| Harvey's Legal Agent Benchmark | — | 0.0 |
| Legal Research Bench | — | 15.9 |
| SAGE | — | 49.9 |
| τ²-Bench - Telecom | 87.0 | — |
| τ³-Banking | — | 14.2 |
| ARC-AGI-1 | — | 65.3 |
| ARC-AGI-2 | — | 11.8 |
| Vending-Bench 2 | — | 1198 |
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.
M2.1: Listed text rates; pricing tier not confirmed.
Kimi K2.5: Listed text rates; pricing tier not confirmed.
| Price / 1M tokens | M2.1 | Kimi K2.5 | 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: M2.1 2026-11-14; Kimi K2.5 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.
| Attribute | M2.1 | Kimi K2.5 |
|---|---|---|
| Availability | Available | Retired |
| Provider | MiniMaxAI | Moonshot AI |
| Model type | Chat model | Multimodal model |
| First release | December 23, 2025 | January 27, 2026 |
| Context window | 200K tokens | 256K tokens |
| Max output | 128K tokens | 16K tokens |
| Total parameters | 230B | 1.0T |
| Active parameters | 10B | 32B |
| Architecture | Mixture of Experts (MoE) | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual |
| Knowledge cutoff | — | 2024-04 |
| Input modalities | Text | Text, Image |
| Output modalities | Text | Text |
| Weights | Open weights | Open weights |
| Licence | Modified MIT | Modified MIT License |
| Commercial use | Commercial use allowed | Commercial use allowed |
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 M2.1 or Kimi K2.5? 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.