Model comparison
Kimi K3 vs Qwen2.5 72B Instruct
Kimi K3 is the stronger model overall, scoring 59.5 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 2.4× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 24 shared benchmarks.
Alibaba (Qwen)Summary
- They share 24 benchmarks with published results for both. Kimi K3 scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K3 leads 74.2 to 19.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 97.2% for Kimi K3 and 8.1% for Qwen2.5 72B Instruct.
- Qwen2.5 72B Instruct is cheaper at $1.40 / $5.60 per million input/output tokens, against $3 / $15 for Kimi K3.
- Kimi K3 accepts more context: 1.05M tokens versus 131K.
Side by side
| Kimi K3 | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | Moonshot AI | Alibaba (Qwen) |
| Noometry Index | 59.5 | 31.9 |
| Released | 2026-07-16 | 2024-09 |
| Weights | Open | Open |
| Context window | 1.05M | 131K |
| Max output | 1.05M | 8K |
| Input $ / M tokens | $3 | $1.40 |
| Output $ / M tokens | $15 | $5.60 |
| Results tracked | 53 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
Kimi K3: 61.0 (#10), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 82.6% | 16% |
| LMArena Coding | 1508 | 1292 |
| DeepSWE | 68.5% | — |
| FrontierCode | 44.2% | — |
| LMArena WebDev | 1654 | — |
| FrontierSWE | 25.9% | — |
| SciCode | 59.5% | — |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
| ALE-Bench | 1,524 | — |
Agentic & Tool Use Kimi K3 leads
Kimi K3: 41.8 (#20), Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| APEX-Agents | 50.6% | — |
| TheAgentCompany | — | 5.7% |
| τ²-bench Banking | 37.1% | — |
| PostTrainBench | 32% | — |
| BALROG | — | 16.2% |
| GBAEval | 48.3% | — |
| GDP.pdf | 19% | — |
| METR Time Horizons | — | 35.8% |
| Vending-Bench 2 | 5,165 | — |
Reasoning Kimi K3 leads
Kimi K3: 63.0 (#17), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1496 | 1271 |
| DTBench | 91.2% | 62.9% |
| LMCA | 52.7% | 13.4% |
| Epoch Capabilities Index | 157.45 | 129 |
| ForecastBench | 61.1 | 57.5 |
| ARC-AGI-2 | 60.4% | — |
| SimpleBench | 60.7% | — |
| NYT Connections (extended) | 93.6% | — |
| ARC-AGI-1 | 94.5% | — |
| CritPt | 23.4% | — |
| Chess Puzzles | 39% | — |
| Mystery Game Puzzles | 26% | — |
| Surface Evolver Bench | 95% | — |
| BIG-Bench Hard | — | 79.8% |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math Kimi K3 leads
Kimi K3: 74.2 (#16), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 97.2% | 8.1% |
| LMArena Math | 1491 | 1283 |
| FrontierMath (Tiers 1-3) | 72.2% | — |
| FrontierMath Tier 4 | 39% | — |
| MathArena Final-Answer Competitions | 87.8% | — |
| ProofBench | 87% | — |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge Kimi K3 leads
Kimi K3: 63.2 (#21), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 93.1% | 49.1% |
| LMArena Expert | 1521 | 1245 |
| SimpleQA Verified | 50.6% | — |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multimodal Not comparable
Kimi K3: 37.8 (#70), Qwen2.5 72B Instruct: —
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| Blueprint-Bench 2 | 29.5% | — |
| Furniture Assembly | 34.2% | — |
Multilingual Kimi K3 leads
Kimi K3: 56.3 (#21), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1466 | 1252 |
| LMArena Chinese | 1529 | 1272 |
| LMArena French | 1491 | 1280 |
| LMArena German | 1488 | 1234 |
| LMArena Japanese | 1487 | 1180 |
| LMArena Korean | 1458 | 1188 |
| LMArena Russian | 1482 | 1264 |
| LMArena Spanish | 1472 | 1256 |
Instruction Following Kimi K3 leads
Kimi K3: 77.7 (#14), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1483 | 1254 |
| IFEval | — | 80.6% |
Long Context Kimi K3 leads
Kimi K3: 45.8 (#29), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1494 | 1282 |
Writing & Preference Kimi K3 leads
Kimi K3: 76.6 (#4), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | Kimi K3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1476 | 1269 |
| LMArena Creative Writing | 1454 | 1221 |
| LMArena Multi-Turn | 1488 | 1272 |
| EQ-Bench Creative Writing | 2082 | — |
| WildBench | — | 80.2% |
| EQ-Bench 4 | 1339 | — |
Frequently asked questions
Is Kimi K3 better than Qwen2.5 72B Instruct?
Kimi K3 is the stronger model overall, scoring 59.5 to 31.9 on the Noometry Index. Qwen2.5 72B Instruct costs 2.4× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, Kimi K3 or Qwen2.5 72B Instruct?
Qwen2.5 72B Instruct is cheaper. It lists at $1.40 per million input tokens and $5.60 per million output tokens; Kimi K3 lists at $3 and $15.
Is Kimi K3 or Qwen2.5 72B Instruct better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
Kimi K3 does, with 1.05M tokens against 131K.
How many benchmarks do Kimi K3 and Qwen2.5 72B Instruct share?
24 benchmarks have published results for both models. Kimi K3 has 53 scored results on Noometry and Qwen2.5 72B Instruct has 43.