Model comparison
DeepSeek-V2.5 (Sep 2024) vs DeepSeek-V3.2-Speciale
DeepSeek-V3.2-Speciale is the stronger model overall, scoring 39.1 to 37.0 on the Noometry Index.
Last verified . 0 shared benchmarks.
DeepSeekSummary
- The widest gap is in coding, where DeepSeek-V3.2-Speciale leads 39.9 to 31.3.
Side by side
| DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 37.0 | 39.1 |
| Released | 2024-09-06 | 2025-12-01 |
| Weights | Open | Open |
| Context window | — | 128K |
| Max output | — | 128K |
| Input $ / M tokens | — | $0.58 |
| Output $ / M tokens | — | $1.68 |
| Results tracked | 22 | 3 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Speciale leads
DeepSeek-V2.5 (Sep 2024): 31.3 (#274), DeepSeek-V3.2-Speciale: 39.9 (#136)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 46.7% |
| BigCodeBench Instruct | 48.6% | — |
| LMArena Coding | 1310 | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Reasoning DeepSeek-V3.2-Speciale leads
DeepSeek-V2.5 (Sep 2024): 24.5 (#145), DeepSeek-V3.2-Speciale: 32.1 (#65)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| SimpleBench | — | 52.6% |
| LMArena Hard Prompts | 1289 | — |
Math Not comparable
DeepSeek-V2.5 (Sep 2024): 33.3 (#177), DeepSeek-V3.2-Speciale: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Math | 1289 | — |
Knowledge Not comparable
DeepSeek-V2.5 (Sep 2024): 34.2 (#190), DeepSeek-V3.2-Speciale: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Expert | 1266 | — |
Multilingual Not comparable
DeepSeek-V2.5 (Sep 2024): 42.7 (#188), DeepSeek-V3.2-Speciale: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Non-English | 1273 | — |
| LMArena Chinese | 1317 | — |
| LMArena French | 1288 | — |
| LMArena German | 1258 | — |
| LMArena Japanese | 1228 | — |
| LMArena Korean | 1209 | — |
| LMArena Russian | 1289 | — |
| LMArena Spanish | 1248 | — |
Instruction Following Not comparable
DeepSeek-V2.5 (Sep 2024): 67.7 (#189), DeepSeek-V3.2-Speciale: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Instruction Following | 1280 | — |
Long Context Not comparable
DeepSeek-V2.5 (Sep 2024): 39.5 (#169), DeepSeek-V3.2-Speciale: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Longer Query | 1301 | — |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 50.0 (#182), DeepSeek-V3.2-Speciale: 46.1 (#217)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Speciale |
|---|---|---|
| LMArena Text | 1294 | — |
| LMArena Creative Writing | 1285 | — |
| EQ-Bench Creative Writing | — | 1276 |
| LMArena Multi-Turn | 1297 | — |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than DeepSeek-V3.2-Speciale?
DeepSeek-V3.2-Speciale is the stronger model overall, scoring 39.1 to 37.0 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or DeepSeek-V3.2-Speciale better for coding?
DeepSeek-V3.2-Speciale scores higher on coding benchmarks: 39.9 versus 31.3 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and DeepSeek-V3.2-Speciale share?
0 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and DeepSeek-V3.2-Speciale has 3.