Model comparison
DeepSeek-V3.2-Speciale vs Kimi K2.5
Kimi K2.5 is the stronger model overall, scoring 48.1 to 39.7 on the Noometry Index.
Last verified . 3 shared benchmarks.
DeepSeekSummary
- They share 3 benchmarks with published results for both. DeepSeek-V3.2-Speciale scores higher in 1 category and Kimi K2.5 in 2 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Kimi K2.5 leads 65.1 to 46.0.
- The biggest single-benchmark swing is SimpleBench: 52.6% for DeepSeek-V3.2-Speciale and 46.8% for Kimi K2.5.
- DeepSeek-V3.2-Speciale is cheaper at $0.58 / $1.68 per million input/output tokens, against $0.45 / $2.25 for Kimi K2.5.
- Kimi K2.5 accepts more context: 262K tokens versus 128K.
Side by side
| DeepSeek-V3.2-Speciale | Kimi K2.5 | |
|---|---|---|
| Provider | DeepSeek | Moonshot AI |
| Noometry Index | 39.7 | 48.1 |
| Released | 2025-12-01 | 2026-01-27 |
| Weights | Open | Open |
| Context window | 128K | 262K |
| Max output | 128K | 262K |
| Input $ / M tokens | $0.58 | $0.45 |
| Output $ / M tokens | $1.68 | $2.25 |
| Results tracked | 3 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
DeepSeek-V3.2-Speciale: 40.4 (#140), Kimi K2.5: 48.8 (#53)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| WeirdML | 46.7% | 45.6% |
| SWE-bench Verified | — | 73.8% |
| SWE-bench Verified (bash only) | — | 70.8% |
| LMArena WebDev | — | 1437 |
| SWE-bench Multilingual | — | 67.3% |
| SciCode | — | 49% |
| LMArena Coding | — | 1474 |
| ALE-Bench | — | 821.65 |
Agentic & Tool Use Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 34.2 (#48)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | — | 43.2% |
| OSWorld | — | 63.3% |
| Vending-Bench 2 | — | 1,198 |
Reasoning DeepSeek-V3.2-Speciale leads
DeepSeek-V3.2-Speciale: 32.9 (#73), Kimi K2.5: 31.2 (#80)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| SimpleBench | 52.6% | 46.8% |
| ARC-AGI-2 | — | 11.8% |
| Kagi LLM Benchmark | — | 78.5% |
| NYT Connections (extended) | — | 69.9% |
| ARC-AGI-1 | — | 65.3% |
| CritPt | — | 3.1% |
| Chess Puzzles | — | 12% |
| EnigmaEval | — | 3.4% |
| Thematic Generalization | — | 69.4% |
| LMArena Hard Prompts | — | 1453 |
| Epoch Capabilities Index | — | 148.03 |
Math Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 51.8 (#53)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| MathArena Final-Answer Competitions | — | 62.3% |
| OTIS Mock AIME 2024-2025 | — | 92.2% |
| LMArena Math | — | 1470 |
| FrontierMath (Feb 2025 set) | — | 27.9% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 53.6 (#56)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | — | 87.6% |
| Humanity's Last Exam | — | 24.4% |
| SimpleQA Verified | — | 34.3% |
| Vectara Hallucination Rate | — | 14.2% |
| LMArena Expert | — | 1466 |
Multimodal Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 41.1 (#39)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| LMArena Vision | — | 1269 |
| LMArena Document | — | 1430 |
Multilingual Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 53.9 (#53)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | — | 1433 |
| LMArena Chinese | — | 1495 |
| LMArena French | — | 1454 |
| LMArena German | — | 1441 |
| LMArena Japanese | — | 1421 |
| LMArena Korean | — | 1410 |
| LMArena Russian | — | 1435 |
| LMArena Spanish | — | 1450 |
Instruction Following Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 75.3 (#64)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | — | 1431 |
Long Context Not comparable
DeepSeek-V3.2-Speciale: —, Kimi K2.5: 52.1 (#7)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |
| LMArena Longer Query | — | 1445 |
Writing & Preference Kimi K2.5 leads
DeepSeek-V3.2-Speciale: 46.0 (#222), Kimi K2.5: 65.1 (#53)
| Benchmark | DeepSeek-V3.2-Speciale | Kimi K2.5 |
|---|---|---|
| EQ-Bench Creative Writing | 1276 | 1579 |
| LMArena Text | — | 1445 |
| LMArena Creative Writing | — | 1423 |
| LMArena Multi-Turn | — | 1444 |
Frequently asked questions
Is DeepSeek-V3.2-Speciale better than Kimi K2.5?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 39.7 on the Noometry Index.
Which is cheaper, DeepSeek-V3.2-Speciale or Kimi K2.5?
DeepSeek-V3.2-Speciale is cheaper. It lists at $0.58 per million input tokens and $1.68 per million output tokens; Kimi K2.5 lists at $0.45 and $2.25.
Is DeepSeek-V3.2-Speciale or Kimi K2.5 better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.8 versus 40.4 in the Noometry coding category.
Which has the bigger context window?
Kimi K2.5 does, with 262K tokens against 128K.
How many benchmarks do DeepSeek-V3.2-Speciale and Kimi K2.5 share?
3 benchmarks have published results for both models. DeepSeek-V3.2-Speciale has 3 scored results on Noometry and Kimi K2.5 has 51.