Model comparison
DeepSeek-V3.1 vs Kimi K2.6
Kimi K2.6 is the stronger model overall, scoring 47.5 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 4.0× less per token, which makes it the better buy when Kimi K2.6's lead doesn't matter for your workload.
Last verified . 23 shared benchmarks.
DeepSeekSummary
- They share 23 benchmarks with published results for both. DeepSeek-V3.1 scores higher in 0 categories and Kimi K2.6 in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2.6 leads 56.8 to 36.1.
- The biggest single-benchmark swing is WeirdML: 38.4% for DeepSeek-V3.1 and 55.9% for Kimi K2.6.
- DeepSeek-V3.1 is cheaper at $0.25 / $0.95 per million input/output tokens, against $0.95 / $4 for Kimi K2.6.
- Kimi K2.6 accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3.1 | Kimi K2.6 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 42.2 | 47.5 |
| Released | 2025-08-21 | 2026-04-20 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 8K | 262K |
| Input $ / M tokens | $0.25 | $0.95 |
| Output $ / M tokens | $0.95 | $4 |
| Results tracked | 27 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.6 leads
DeepSeek-V3.1: 39.8 (#140), Kimi K2.6: 50.5 (#41)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| WeirdML | 38.4% | 55.9% |
| LMArena Coding | 1417 | 1488 |
| SWE-bench Verified | — | 76.7% |
| LMArena WebDev | — | 1509 |
| SciCode | — | 53.5% |
| ALE-Bench | — | 1,093 |
Agentic & Tool Use Not comparable
DeepSeek-V3.1: —, Kimi K2.6: 21.8 (#128)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| OSWorld 2.0 | — | 4.6% |
| ExploitBench | — | 18.4% |
| GBAEval | — | 0.9% |
| GDP.pdf | — | 12% |
| Vending-Bench 2 | — | 6,205 |
Reasoning Kimi K2.6 leads
DeepSeek-V3.1: 27.3 (#109), Kimi K2.6: 39.6 (#51)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Hard Prompts | 1417 | 1470 |
| DTBench | 82.7% | 90.9% |
| LMCA | 24.3% | 37.3% |
| Epoch Capabilities Index | 139.92 | 151.03 |
| SimpleBench | 40% | — |
| Kagi LLM Benchmark | 53.2% | — |
| NYT Connections (extended) | — | 87.2% |
| CritPt | — | 8% |
| Chess Puzzles | — | 26% |
| EBR-Bench | — | 2.4% |
| Mystery Game Puzzles | — | 18% |
| ForecastBench | 58 | — |
Math Kimi K2.6 leads
DeepSeek-V3.1: 36.1 (#125), Kimi K2.6: 56.8 (#34)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Math | 1420 | 1475 |
| FrontierMath (Tiers 1-3) | — | 57.2% |
| FrontierMath Tier 4 | — | 25.6% |
| MathArena Final-Answer Competitions | — | 72.9% |
| OTIS Mock AIME 2024-2025 | — | 96.1% |
| ProofBench | — | 16% |
| FrontierMath (Feb 2025 set) | — | 39% |
| FrontierMath Tier 4 (v1) | — | 14.6% |
Knowledge Kimi K2.6 leads
DeepSeek-V3.1: 43.0 (#86), Kimi K2.6: 53.9 (#48)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| Vectara Hallucination Rate | 5.5% | 10.8% |
| LMArena Expert | 1403 | 1491 |
| GPQA Diamond | — | 90.8% |
| SimpleQA Verified | — | 34.9% |
Multimodal Not comparable
DeepSeek-V3.1: —, Kimi K2.6: 31.8 (#95)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Vision | — | 1282 |
| Blueprint-Bench 2 | — | 3.9% |
| Furniture Assembly | — | 21.7% |
| LMArena Document | — | 1451 |
Multilingual Kimi K2.6 leads
DeepSeek-V3.1: 51.8 (#100), Kimi K2.6: 55.1 (#35)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Non-English | 1400 | 1446 |
| LMArena Chinese | 1470 | 1521 |
| LMArena French | 1447 | 1471 |
| LMArena German | 1412 | 1449 |
| LMArena Japanese | 1374 | 1442 |
| LMArena Korean | 1337 | 1428 |
| LMArena Russian | 1405 | 1446 |
| LMArena Spanish | 1429 | 1463 |
Instruction Following Kimi K2.6 leads
DeepSeek-V3.1: 74.1 (#105), Kimi K2.6: 76.5 (#40)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Instruction Following | 1400 | 1451 |
Long Context Kimi K2.6 leads
DeepSeek-V3.1: 36.3 (#227), Kimi K2.6: 44.9 (#50)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Longer Query | 1422 | 1468 |
| Fiction.LiveBench | 52.8% | — |
Writing & Preference Kimi K2.6 leads
DeepSeek-V3.1: 60.5 (#93), Kimi K2.6: 68.7 (#25)
| Benchmark | DeepSeek-V3.1 | Kimi K2.6 |
|---|---|---|
| LMArena Text | 1420 | 1455 |
| LMArena Creative Writing | 1401 | 1434 |
| EQ-Bench Creative Writing | 1436 | 1725 |
| LMArena Multi-Turn | 1408 | 1453 |
| EQ-Bench 4 | — | 1202 |
Frequently asked questions
Is DeepSeek-V3.1 better than Kimi K2.6?
Kimi K2.6 is the stronger model overall, scoring 47.5 to 42.2 on the Noometry Index. DeepSeek-V3.1 costs 4.0× less per token, which makes it the better buy when Kimi K2.6's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.1 or Kimi K2.6?
DeepSeek-V3.1 is cheaper. It lists at $0.25 per million input tokens and $0.95 per million output tokens; Kimi K2.6 lists at $0.95 and $4.
Is DeepSeek-V3.1 or Kimi K2.6 better for coding?
Kimi K2.6 scores higher on coding benchmarks: 50.5 versus 39.8 in the Noometry coding category.
Which has the bigger context window?
Kimi K2.6 does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3.1 and Kimi K2.6 share?
23 benchmarks have published results for both models. DeepSeek-V3.1 has 27 scored results on Noometry and Kimi K2.6 has 51.