Model comparison
DeepSeek-V3.2-Exp vs Kimi K2.5
Kimi K2.5 is the stronger model overall, scoring 48.1 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 3.1× less per token, which makes it the better buy when Kimi K2.5's lead doesn't matter for your workload.
Last verified . 42 shared benchmarks.
DeepSeekSummary
- They share 42 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 0 categories and Kimi K2.5 in 9 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2.5 leads 51.8 to 41.7.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 69.9% for Kimi K2.5.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $0.45 / $2.25 for Kimi K2.5.
- Kimi K2.5 accepts more context: 262K tokens versus 164K.
Side by side
| DeepSeek-V3.2-Exp | Kimi K2.5 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 44.3 | 48.1 |
| Released | 2025-09-29 | 2026-01-27 |
| Weights | Open | Open |
| Context window | 164K | 262K |
| Max output | 66K | 262K |
| Input $ / M tokens | $0.26 | $0.45 |
| Output $ / M tokens | $0.38 | $2.25 |
| Results tracked | 49 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
DeepSeek-V3.2-Exp: 46.5 (#65), Kimi K2.5: 48.8 (#53)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 70.8% |
| LMArena WebDev | 1362 | 1437 |
| SWE-bench Multilingual | 59% | 67.3% |
| SciCode | 38.9% | 49% |
| WeirdML | 39.5% | 45.6% |
| LMArena Coding | 1454 | 1474 |
| SWE-bench Verified | — | 73.8% |
| Aider Polyglot | 74.2% | — |
| ALE-Bench | — | 821.65 |
Agentic & Tool Use Kimi K2.5 leads
DeepSeek-V3.2-Exp: 32.7 (#59), Kimi K2.5: 34.2 (#48)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | 39.6% | 43.2% |
| Vending-Bench 2 | 1,034 | 1,198 |
| APEX-Agents | 21.3% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| OSWorld | — | 63.3% |
Reasoning Kimi K2.5 leads
DeepSeek-V3.2-Exp: 22.1 (#208), Kimi K2.5: 31.2 (#80)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| ARC-AGI-2 | 4% | 11.8% |
| Kagi LLM Benchmark | 52.2% | 78.5% |
| NYT Connections (extended) | 36.7% | 69.9% |
| ARC-AGI-1 | 57% | 65.3% |
| CritPt | 2.9% | 3.1% |
| Chess Puzzles | 14% | 12% |
| Thematic Generalization | 65% | 69.4% |
| LMArena Hard Prompts | 1434 | 1453 |
| Epoch Capabilities Index | 146.27 | 148.03 |
| SimpleBench | — | 46.8% |
| EnigmaEval | — | 3.4% |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
Math Kimi K2.5 leads
DeepSeek-V3.2-Exp: 41.7 (#87), Kimi K2.5: 51.8 (#53)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 62.3% |
| OTIS Mock AIME 2024-2025 | 87.8% | 92.2% |
| LMArena Math | 1435 | 1470 |
| FrontierMath (Feb 2025 set) | 22.1% | 27.9% |
| FrontierMath Tier 4 (v1) | 2.1% | 4.2% |
| ProofBench | 8% | — |
Knowledge Kimi K2.5 leads
DeepSeek-V3.2-Exp: 51.7 (#66), Kimi K2.5: 53.6 (#56)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | 83.4% | 87.6% |
| Vectara Hallucination Rate | 5.3% | 14.2% |
| LMArena Expert | 1436 | 1466 |
| Humanity's Last Exam | — | 24.4% |
| SimpleQA Verified | — | 34.3% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Kimi K2.5: 41.1 (#39)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| LMArena Vision | — | 1269 |
| LMArena Document | — | 1430 |
Multilingual Kimi K2.5 leads
DeepSeek-V3.2-Exp: 52.2 (#90), Kimi K2.5: 53.9 (#53)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1409 | 1433 |
| LMArena Chinese | 1461 | 1495 |
| LMArena French | 1433 | 1454 |
| LMArena German | 1440 | 1441 |
| LMArena Japanese | 1374 | 1421 |
| LMArena Korean | 1371 | 1410 |
| LMArena Russian | 1424 | 1435 |
| LMArena Spanish | 1440 | 1450 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), Kimi K2.5: 75.3 (#64)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1413 | 1431 |
Long Context Kimi K2.5 leads
DeepSeek-V3.2-Exp: 47.6 (#16), Kimi K2.5: 52.1 (#7)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| Fiction.LiveBench | 83.3% | 86.1% |
| CL-bench | 13.2% | 19.3% |
| CL-bench Life | 9.5% | 13.2% |
| LMArena Longer Query | 1428 | 1445 |
Writing & Preference Kimi K2.5 leads
DeepSeek-V3.2-Exp: 62.4 (#77), Kimi K2.5: 65.1 (#53)
| Benchmark | DeepSeek-V3.2-Exp | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1425 | 1445 |
| LMArena Creative Writing | 1403 | 1423 |
| EQ-Bench Creative Writing | 1515 | 1579 |
| LMArena Multi-Turn | 1427 | 1444 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Kimi K2.5?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 44.3 on the Noometry Index. DeepSeek-V3.2-Exp costs 3.1× less per token, which makes it the better buy when Kimi K2.5's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Exp or Kimi K2.5?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Kimi K2.5 lists at $0.45 and $2.25.
Is DeepSeek-V3.2-Exp or Kimi K2.5 better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.8 versus 46.5 in the Noometry coding category.
Which has the bigger context window?
Kimi K2.5 does, with 262K tokens against 164K.
How many benchmarks do DeepSeek-V3.2-Exp and Kimi K2.5 share?
42 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Kimi K2.5 has 51.