Compare / head-to-head
Kimi K2.5vs
MiniMax M2.1
Kimi K2.5 leads 3 of 3 shared benchmarks. Kimi K2.5 has the larger context window (256K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
3 – 0
Kimi K2.5 leads
Cheaper per token
—
list price, input + output
Larger context
Kimi K2.5
256K tokens
Providers
2 providers
Moonshot AI · MiniMax
Kimi K2.5
Moonshot AI · released 2026-01-27
textvisionvideo
- Context
- 256K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 26 · 14 core
MiniMax M2.1
MiniMax · released 2025-12-22
text
Quality
Benchmark matrix
3 shared · 23 only Kimi K2.5 · 10 only MiniMax M2.1| Benchmark | Kimi K2.5 | MiniMax M2.1 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 3 | ||||
| SWE-bench Verified | 76.8% ↗ | 74% ↗ | +2.8 pt | Kimi K2.5 |
| BrowseComp | 60.6% ↗ | 47.4% ↗ | +13.2 pt | Kimi K2.5 |
| SimpleBench (SimpleBench) | 46.8% ↗ | 34.7% ↗ | +12.1 pt | Kimi K2.5 |
| Only Kimi K2.5 reports · 23 | ||||
| MMLU-Pro | 87.1% ↗ | not reported | — | — |
| AIME 2025 | 96.1% ↗ | not reported | — | — |
| LMArena Elo | 1450.1 ↗ | not reported | — | — |
| GPQA-Diamond | 87.6% ↗ | not reported | — | — |
| HLE-Full | 30.1% ↗ | not reported | — | — |
| HLE-Full (w/ tools) | 50.2% ↗ | not reported | — | — |
| HMMT 2025 (Feb) | 95.4% ↗ | not reported | — | — |
| IMO-AnswerBench | 81.8% ↗ | not reported | — | — |
| LiveCodeBench (v6) | 85% ↗ | not reported | — | — |
| LMArena Vision (LMArena) | 1252.2 ↗ | not reported | — | — |
| LMArena WebDev (LMArena) | 1435.8 ↗ | not reported | — | — |
| MMMU-Pro | 78.5% ↗ | not reported | — | — |
| OSWorld-Verified (XLANG) | 63.3% ↗ | not reported | — | — |
| SAGE (Vals AI) | 49.9% ↗ | not reported | — | — |
| SimpleQA Verified | 34.3% ↗epoch run | not reported | — | — |
| SWE-Bench Multilingual | 73% ↗ | not reported | — | — |
| SWE-Bench Pro | 50.7% ↗ | not reported | — | — |
| SWE-bench Verified (Epoch AI run) | 73.8% ↗epoch run | not reported | — | — |
| Terminal Bench 2.0 | 50.8% ↗ | not reported | — | — |
| Toolathlon-Verified (HKUST) | 33% ↗ | not reported | — | — |
| Vending-Bench 2 (Andon Labs) | 1198.46 ↗ | not reported | — | — |
| VideoMMMU | 86.6% ↗ | not reported | — | — |
| WeirdML (Håvard Tveit Ihle) | 45.6% ↗ | not reported | — | — |
| Only MiniMax M2.1 reports · 10 | ||||
| BrowseComp (context management) | not reported | 62% ↗ | — | — |
| Multi-SWE-bench | not reported | 49.4% ↗ | — | — |
| OctoCodingbench | not reported | 26.1% ↗ | — | — |
| SWE-bench Multilingual | not reported | 72.5% ↗ | — | — |
| SWE-bench Verified (Droid) | not reported | 71.3% ↗ | — | — |
| SWE-bench Verified (mini-swe-agent) | not reported | 67% ↗ | — | — |
| SWT-bench | not reported | 69.3% ↗ | — | — |
| Terminal-bench 2.0 | not reported | 47.9% ↗ | — | — |
| Toolathlon | not reported | 43.5% ↗ | — | — |
| VIBE (Average) | not reported | 88.6% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Kimi K2.5 minus MiniMax M2.1 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Kimi K2.5 | MiniMax M2.1 | Edge |
|---|---|---|---|
| Context window | 256K tokens | 205K tokens | Kimi K2.5 |
| Max output | — | — | — |
| Input price / 1M | — | $0.3 ↗ | — |
| Output price / 1M | — | $1.2 ↗ | — |
| Cached input / 1M | — | $0.03 ↗ | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | $1.5 | — |
| Modalities | text · vision · video | text | Kimi K2.5 |
| Released | 2026-01-27 | 2025-12-22 | — |
| Cited benchmark scores | 26 | 13 | — |
Reliability
Provider status
Live from /statusMore matchups
Kimi K2.5 vs …
Models sharing the most benchmarksMore matchups
MiniMax M2.1 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Kimi K2.5 and MiniMax M2.1 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.