Model comparison
DeepSeek-V3.2-Exp vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 21× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 35 shared benchmarks.
DeepSeekSummary
- They share 35 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 1 category and Kimi K3 in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K3 leads 63.0 to 22.1.
- The biggest single-benchmark swing is ProofBench: 8% for DeepSeek-V3.2-Exp and 87% for Kimi K3.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $3 / $15 for Kimi K3.
- Kimi K3 accepts more context: 1.05M tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | Kimi K3 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 44.3 | 59.5 |
| Released | 2025-09-29 | 2026-07-16 |
| Weights | Open | Open |
| Context window | 164K | 1.05M |
| Max output | 66K | 1.05M |
| Input $ / M tokens | $0.26 | $3 |
| Output $ / M tokens | $0.38 | $15 |
| Results tracked | 49 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
DeepSeek-V3.2-Exp: 46.5 (#65), Kimi K3: 61.0 (#10)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| LMArena WebDev | 1362 | 1654 |
| SciCode | 38.9% | 59.5% |
| WeirdML | 39.5% | 82.6% |
| LMArena Coding | 1454 | 1508 |
| DeepSWE | — | 68.5% |
| FrontierCode | — | 44.2% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| FrontierSWE | — | 25.9% |
| ALE-Bench | — | 1,524 |
Agentic & Tool Use Kimi K3 leads
DeepSeek-V3.2-Exp: 32.7 (#59), Kimi K3: 41.8 (#20)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| APEX-Agents | 21.3% | 50.6% |
| Vending-Bench 2 | 1,034 | 5,165 |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 32% |
| GBAEval | — | 48.3% |
| GDP.pdf | — | 19% |
Reasoning Kimi K3 leads
DeepSeek-V3.2-Exp: 22.1 (#208), Kimi K3: 63.0 (#17)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 4% | 60.4% |
| NYT Connections (extended) | 36.7% | 93.6% |
| ARC-AGI-1 | 57% | 94.5% |
| CritPt | 2.9% | 23.4% |
| Chess Puzzles | 14% | 39% |
| LMArena Hard Prompts | 1434 | 1496 |
| DTBench | 87.7% | 91.2% |
| LMCA | 29.1% | 52.7% |
| Epoch Capabilities Index | 146.27 | 157.45 |
| SimpleBench | — | 60.7% |
| Kagi LLM Benchmark | 52.2% | — |
| Thematic Generalization | 65% | — |
| Mystery Game Puzzles | — | 26% |
| Surface Evolver Bench | — | 95% |
| ForecastBench | — | 61.1 |
Math Kimi K3 leads
DeepSeek-V3.2-Exp: 41.7 (#87), Kimi K3: 74.2 (#16)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 87.8% |
| OTIS Mock AIME 2024-2025 | 87.8% | 97.2% |
| ProofBench | 8% | 87% |
| LMArena Math | 1435 | 1491 |
| FrontierMath (Tiers 1-3) | — | 72.2% |
| FrontierMath Tier 4 | — | 39% |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Kimi K3 leads
DeepSeek-V3.2-Exp: 51.7 (#66), Kimi K3: 63.2 (#21)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| GPQA Diamond | 83.4% | 93.1% |
| LMArena Expert | 1436 | 1521 |
| SimpleQA Verified | — | 50.6% |
| Vectara Hallucination Rate | 5.3% | — |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Kimi K3: 37.8 (#70)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| Blueprint-Bench 2 | — | 29.5% |
| Furniture Assembly | — | 34.2% |
Multilingual Kimi K3 leads
DeepSeek-V3.2-Exp: 52.2 (#90), Kimi K3: 56.3 (#21)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1409 | 1466 |
| LMArena Chinese | 1461 | 1529 |
| LMArena French | 1433 | 1491 |
| LMArena German | 1440 | 1488 |
| LMArena Japanese | 1374 | 1487 |
| LMArena Korean | 1371 | 1458 |
| LMArena Russian | 1424 | 1482 |
| LMArena Spanish | 1440 | 1472 |
Instruction Following Kimi K3 leads
DeepSeek-V3.2-Exp: 74.5 (#93), Kimi K3: 77.7 (#14)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1483 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Kimi K3: 45.8 (#29)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1428 | 1494 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Kimi K3 leads
DeepSeek-V3.2-Exp: 62.4 (#77), Kimi K3: 76.6 (#4)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K3 |
|---|---|---|
| LMArena Text | 1425 | 1476 |
| LMArena Creative Writing | 1403 | 1454 |
| EQ-Bench Creative Writing | 1515 | 2082 |
| LMArena Multi-Turn | 1427 | 1488 |
| EQ-Bench 4 | — | 1339 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 21× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or Kimi K3?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Kimi K3 lists at $3 and $15.
Is DeepSeek-V3.2-Exp or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 46.5 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-V3.2-Exp and Kimi K3 share?
35 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Kimi K3 has 53.