Model comparison
DeepSeek-V2.5 (Sep 2024) vs DeepSeek-V3.2-Exp
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 37.6 on the Noometry Index.
Last verified . 18 shared benchmarks.
DeepSeekSummary
- They share 18 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 1 category and DeepSeek-V3.2-Exp in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V3.2-Exp leads 51.7 to 34.8.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 74.2% for DeepSeek-V3.2-Exp.
Side by side
| DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 37.6 | 44.3 |
| Released | 2024-09-06 | 2025-09-29 |
| Weights | Open | Open |
| Context window | — | 164K |
| Max output | — | 66K |
| Input $ / M tokens | — | $0.26 |
| Output $ / M tokens | — | $0.38 |
| Results tracked | 22 | 49 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), DeepSeek-V3.2-Exp: 46.5 (#65)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| Aider Polyglot | 17.8% | 74.2% |
| LMArena Coding | 1309 | 1454 |
| SWE-bench Verified (bash only) | — | 70% |
| LMArena WebDev | — | 1362 |
| SWE-bench Multilingual | — | 59% |
| SciCode | — | 38.9% |
| WeirdML | — | 39.5% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, DeepSeek-V3.2-Exp: 32.7 (#59)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| Terminal-Bench | — | 39.6% |
| APEX-Agents | — | 21.3% |
| Berkeley Function Calling Leaderboard | — | 56.7% |
| TheAgentCompany | — | 42.9% |
| Vending-Bench 2 | — | 1,034 |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), DeepSeek-V3.2-Exp: 22.1 (#208)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1434 |
| ARC-AGI-2 | — | 4% |
| Kagi LLM Benchmark | — | 52.2% |
| NYT Connections (extended) | — | 36.7% |
| ARC-AGI-1 | — | 57% |
| CritPt | — | 2.9% |
| Chess Puzzles | — | 14% |
| Thematic Generalization | — | 65% |
| DTBench | — | 87.7% |
| LMCA | — | 29.1% |
| Epoch Capabilities Index | — | 146.27 |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), DeepSeek-V3.2-Exp: 41.7 (#87)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Math | 1288 | 1435 |
| MathArena Final-Answer Competitions | — | 57.7% |
| OTIS Mock AIME 2024-2025 | — | 87.8% |
| ProofBench | — | 8% |
| FrontierMath (Feb 2025 set) | — | 22.1% |
| FrontierMath Tier 4 (v1) | — | 2.1% |
Knowledge DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), DeepSeek-V3.2-Exp: 51.7 (#66)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Expert | 1266 | 1436 |
| GPQA Diamond | — | 83.4% |
| Vectara Hallucination Rate | — | 5.3% |
Multilingual DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), DeepSeek-V3.2-Exp: 52.2 (#90)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Non-English | 1273 | 1409 |
| LMArena Chinese | 1318 | 1461 |
| LMArena French | 1289 | 1433 |
| LMArena German | 1258 | 1440 |
| LMArena Japanese | 1228 | 1374 |
| LMArena Korean | 1209 | 1371 |
| LMArena Russian | 1289 | 1424 |
| LMArena Spanish | 1248 | 1440 |
Instruction Following DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), DeepSeek-V3.2-Exp: 74.5 (#93)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Instruction Following | 1280 | 1413 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), DeepSeek-V3.2-Exp: 47.6 (#16)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Longer Query | 1301 | 1428 |
| Fiction.LiveBench | — | 83.3% |
| CL-bench | — | 13.2% |
| CL-bench Life | — | 9.5% |
Writing & Preference DeepSeek-V3.2-Exp leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), DeepSeek-V3.2-Exp: 62.4 (#77)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | DeepSeek-V3.2-Exp |
|---|---|---|
| LMArena Text | 1294 | 1425 |
| LMArena Creative Writing | 1285 | 1403 |
| LMArena Multi-Turn | 1297 | 1427 |
| EQ-Bench Creative Writing | — | 1515 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than DeepSeek-V3.2-Exp?
DeepSeek-V3.2-Exp is the stronger model overall, scoring 44.3 to 37.6 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or DeepSeek-V3.2-Exp better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and DeepSeek-V3.2-Exp share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and DeepSeek-V3.2-Exp has 49.