Model comparison
DeepSeek-R1 vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 42.3 on the Noometry Index. DeepSeek-R1 costs 6.6× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 29 shared benchmarks.
DeepSeekSummary
- They share 29 benchmarks with published results for both. DeepSeek-R1 scores higher in 0 categories and Kimi K3 in 9 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K3 leads 63.0 to 18.6.
- The biggest single-benchmark swing is ARC-AGI-1: 21.2% for DeepSeek-R1 and 94.5% for Kimi K3.
- DeepSeek-R1 is cheaper at $0.50 / $2.15 per million input/output tokens, against $3 / $15 for Kimi K3.
- Kimi K3 accepts more context: 1.05M tokens versus 164K.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-R1 | Kimi K3 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 42.3 | 59.5 |
| Released | 2025-01-20 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 164K | 1.05M |
| Max output | 64K | 1.05M |
| Input $ / M tokens | $0.50 | $3 |
| Output $ / M tokens | $2.15 | $15 |
| Results tracked | 52 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
DeepSeek-R1: 46.3 (#68), Kimi K3: 61.0 (#10)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| SciCode | 35.7% | 59.5% |
| WeirdML | 41.6% | 82.6% |
| LMArena Coding | 1427 | 1508 |
| ALE-Bench | 804.12 | 1,524 |
| DeepSWE | — | 68.5% |
| FrontierCode | — | 44.2% |
| Aider Polyglot | 71.4% | — |
| LMArena WebDev | — | 1654 |
| FrontierSWE | — | 25.9% |
| LiveBench Coding | 66.7% | — |
| AlgoTune | 1.7 | — |
Agentic & Tool Use Kimi K3 leads
DeepSeek-R1: 30.7 (#75), Kimi K3: 41.8 (#20)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| APEX-Agents | — | 50.6% |
| τ²-bench Banking | — | 37.1% |
| DeepResearch Bench | 35.1% | — |
| PostTrainBench | — | 32% |
| BALROG | 34.9% | — |
| GBAEval | — | 48.3% |
| GDP.pdf | — | 19% |
| METR Time Horizons | 53.8% | — |
| Vending-Bench 2 | — | 5,165 |
Reasoning Kimi K3 leads
DeepSeek-R1: 18.6 (#278), Kimi K3: 63.0 (#17)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 1.3% | 60.4% |
| SimpleBench | 40.8% | 60.7% |
| ARC-AGI-1 | 21.2% | 94.5% |
| CritPt | 1.1% | 23.4% |
| LMArena Hard Prompts | 1416 | 1496 |
| Epoch Capabilities Index | 141.29 | 157.45 |
| ForecastBench | 60 | 61.1 |
| Kagi LLM Benchmark | 69.4% | — |
| NYT Connections (extended) | — | 93.6% |
| Chess Puzzles | — | 39% |
| LiveBench Reasoning | 83.2% | — |
| Mystery Game Puzzles | — | 26% |
| DTBench | — | 91.2% |
| LiveBench Data Analysis | 69.8% | — |
| LMCA | — | 52.7% |
| Surface Evolver Bench | — | 95% |
| LiveBench | 71.6% | — |
Math Kimi K3 leads
DeepSeek-R1: 43.8 (#79), Kimi K3: 74.2 (#16)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.4% | 97.2% |
| LMArena Math | 1400 | 1491 |
| FrontierMath (Tiers 1-3) | — | 72.2% |
| FrontierMath Tier 4 | — | 39% |
| MathArena Final-Answer Competitions | — | 87.8% |
| ProofBench | — | 87% |
| Omni-MATH | 42.4% | — |
| LiveBench Math | 80.7% | — |
| MATH Level 5 | 96.6% | — |
Knowledge Kimi K3 leads
DeepSeek-R1: 44.5 (#87), Kimi K3: 63.2 (#21)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| GPQA Diamond | 76.3% | 93.1% |
| LMArena Expert | 1394 | 1521 |
| SimpleQA Verified | — | 50.6% |
| MMLU-Pro | 79.3% | — |
| Confabulations | 12.7% | — |
| Vectara Hallucination Rate | 11.3% | — |
| GPQA (HELM) | 66.6% | — |
Multimodal Not comparable
DeepSeek-R1: —, Kimi K3: 37.8 (#70)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | — | 29.5% |
| Furniture Assembly | — | 34.2% |
Multilingual Kimi K3 leads
DeepSeek-R1: 52.4 (#85), Kimi K3: 56.3 (#21)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1412 | 1466 |
| LMArena Chinese | 1442 | 1529 |
| LMArena French | 1417 | 1491 |
| LMArena German | 1404 | 1488 |
| LMArena Japanese | 1391 | 1487 |
| LMArena Korean | 1360 | 1458 |
| LMArena Russian | 1423 | 1482 |
| LMArena Spanish | 1411 | 1472 |
Instruction Following Kimi K3 leads
DeepSeek-R1: 72.0 (#143), Kimi K3: 77.7 (#14)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1382 | 1483 |
| LiveBench Instruction Following | 80.5% | — |
| IFEval | 78.4% | — |
Long Context Too close to call
DeepSeek-R1: 45.4 (#36), Kimi K3: 45.8 (#29)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1391 | 1494 |
| Fiction.LiveBench | 75% | — |
Writing & Preference Kimi K3 leads
DeepSeek-R1: 61.4 (#88), Kimi K3: 76.6 (#4)
| Benchmark | DeepSeek-R1 | Kimi K3 |
|---|---|---|
| LMArena Text | 1428 | 1476 |
| LMArena Creative Writing | 1405 | 1454 |
| EQ-Bench Creative Writing | 1500 | 2082 |
| LMArena Multi-Turn | 1405 | 1488 |
| Short-Story Creative Writing | 83% | — |
| WildBench | 82.8% | — |
| EQ-Bench 4 | — | 1339 |
| LiveBench Language | 48.5% | — |
Frequently asked questions
Is DeepSeek-R1 better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 42.3 on the Noometry Index. DeepSeek-R1 costs 6.6× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-R1 or Kimi K3?
DeepSeek-R1 is cheaper. It lists at $0.50 per million input tokens and $2.15 per million output tokens; Kimi K3 lists at $3 and $15.
Is DeepSeek-R1 or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 46.3 in the Noometry coding category.
Which has the bigger context window?
Kimi K3 does, with 1.05M tokens against 164K.
How many benchmarks do DeepSeek-R1 and Kimi K3 share?
29 benchmarks have published results for both models. DeepSeek-R1 has 52 scored results on Noometry and Kimi K3 has 53.