Model comparison
DeepSeek-R1 vs Kimi K2 (Jul 2025)
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 41.2 on the Noometry Index.
Last verified . 35 shared benchmarks.
DeepSeekSummary
- They share 35 benchmarks with published results for both. DeepSeek-R1 scores higher in 6 categories and Kimi K2 (Jul 2025) in 3 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-R1 leads 44.5 to 37.3.
- The biggest single-benchmark swing is Omni-MATH: 42.4% for DeepSeek-R1 and 65.4% for Kimi K2 (Jul 2025).
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $0.57 / $2.30 for Kimi K2 (Jul 2025).
- Kimi K2 (Jul 2025) accepts more context: 262K tokens versus 164K.
- Kimi K2 (Jul 2025) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | Kimi K2 (Jul 2025) | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 42.3 | 41.2 |
| Released | 2025-01-20 | 2025-07-12 |
| Weights | Proprietary | Open |
| Context window | 164K | 262K |
| Max output | 64K | 262K |
| Input $ / M tokens | $0.50 | $0.57 |
| Output $ / M tokens | $2.15 | $2.30 |
| Results tracked | 52 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
DeepSeek-R1: 46.3 (#68), Kimi K2 (Jul 2025): 42.4 (#102)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| Aider Polyglot | 71.4% | 59.1% |
| WeirdML | 41.6% | 42.8% |
| LMArena Coding | 1427 | 1399 |
| ALE-Bench | 804.12 | 597.5 |
| SWE-bench Verified (bash only) | — | 63.4% |
| SciCode | 35.7% | — |
| GSO | — | 4.9% |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use Kimi K2 (Jul 2025) leads
DeepSeek-R1: 30.7 (#75), Kimi K2 (Jul 2025): 32.4 (#64)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| METR Time Horizons | 53.8% | 59.2% |
| Terminal-Bench | — | 35.7% |
| Berkeley Function Calling Leaderboard | — | 59.1% |
| DeepResearch Bench | 35.1% | — |
| BALROG | 34.9% | — |
Reasoning Kimi K2 (Jul 2025) leads
DeepSeek-R1: 18.6 (#278), Kimi K2 (Jul 2025): 23.3 (#179)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| SimpleBench | 40.8% | 26.3% |
| Kagi LLM Benchmark | 69.4% | 64.4% |
| LMArena Hard Prompts | 1416 | 1384 |
| Epoch Capabilities Index | 141.29 | 146.01 |
| ForecastBench | 60 | 60.2 |
| ARC-AGI-2 | 1.3% | — |
| ARC-AGI-1 | 21.2% | — |
| CritPt | 1.1% | — |
| LiveBench Reasoning | 83.2% | — |
| LiveBench Data Analysis | 69.8% | — |
| LiveBench | 71.6% | — |
Math DeepSeek-R1 leads
DeepSeek-R1: 43.8 (#79), Kimi K2 (Jul 2025): 42.7 (#83)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| Omni-MATH | 42.4% | 65.4% |
| LMArena Math | 1400 | 1397 |
| OTIS Mock AIME 2024-2025 | 66.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
| FrontierMath (Feb 2025 set) | — | 21.4% |
| FrontierMath Tier 4 (v1) | — | 0% |
Knowledge DeepSeek-R1 leads
DeepSeek-R1: 44.5 (#87), Kimi K2 (Jul 2025): 37.3 (#157)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| MMLU-Pro | 79.3% | 81.9% |
| Confabulations | 12.7% | 20.4% |
| Vectara Hallucination Rate | 11.3% | 17.9% |
| GPQA (HELM) | 66.6% | 65.3% |
| LMArena Expert | 1394 | 1365 |
| GPQA Diamond | 76.3% | — |
Multilingual DeepSeek-R1 leads
DeepSeek-R1: 52.4 (#85), Kimi K2 (Jul 2025): 49.6 (#130)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| LMArena Non-English | 1412 | 1372 |
| LMArena Chinese | 1442 | 1415 |
| LMArena French | 1417 | 1379 |
| LMArena German | 1404 | 1387 |
| LMArena Japanese | 1391 | 1349 |
| LMArena Korean | 1360 | 1325 |
| LMArena Russian | 1423 | 1385 |
| LMArena Spanish | 1411 | 1386 |
Instruction Following Too close to call
DeepSeek-R1: 72.0 (#143), Kimi K2 (Jul 2025): 71.1 (#156)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| IFEval | 78.4% | 85% |
| LMArena Instruction Following | 1382 | 1348 |
| LiveBench Instruction Following | 80.5% | — |
Long Context DeepSeek-R1 leads
DeepSeek-R1: 45.4 (#36), Kimi K2 (Jul 2025): 41.2 (#145)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| Fiction.LiveBench | 75% | 66.7% |
| LMArena Longer Query | 1391 | 1353 |
| CL-bench | — | 17.6% |
Writing & Preference Too close to call
DeepSeek-R1: 61.4 (#88), Kimi K2 (Jul 2025): 62.3 (#78)
| Benchmark | DeepSeek-R1 | Kimi K2 (Jul 2025) |
|---|---|---|
| LMArena Text | 1428 | 1380 |
| LMArena Creative Writing | 1405 | 1350 |
| Short-Story Creative Writing | 83% | 85.6% |
| EQ-Bench Creative Writing | 1500 | 1666 |
| WildBench | 82.8% | 86.2% |
| LMArena Multi-Turn | 1405 | 1371 |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than Kimi K2 (Jul 2025)?
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 41.2 on the Noometry Index.
Which is cheaper, DeepSeek-R1 or Kimi K2 (Jul 2025)?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; Kimi K2 (Jul 2025) lists at $0.57 and $2.30.
Is DeepSeek-R1 or Kimi K2 (Jul 2025) better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 46.3 versus 42.4 in the Noometry coding category.
Which has the bigger context window?
Kimi K2 (Jul 2025) does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-R1 and Kimi K2 (Jul 2025) share?
35 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Kimi K2 (Jul 2025) has 42.