Model comparison
DeepSeek-V3.2-Exp vs Gemini 3 Pro
Gemini 3 Pro is the stronger model overall, scoring 54.7 to 44.9 on the Noometry Index.
Last verified . 40 shared benchmarks.
DeepSeek
GoogleSummary
- They share 40 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 1 category and Gemini 3 Pro in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Gemini 3 Pro leads 52.5 to 22.0.
- The biggest single-benchmark swing is NYT Connections (extended): 36.7% for DeepSeek-V3.2-Exp and 94.4% for Gemini 3 Pro.
- DeepSeek-V3.2-Exp has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3.2-Exp | Gemini 3 Pro | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 44.9 | 54.7 |
| Released | 2025-09-29 | 2025-11-18 |
| Weights | Open | Proprietary |
| Context window | 164K | — |
| Max output | 66K | — |
| Input $ / M tokens | $0.26 | — |
| Output $ / M tokens | $0.38 | — |
| Results tracked | 48 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 46.4 (#61), Gemini 3 Pro: 52.1 (#33)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| SWE-bench Verified (bash only) | 70% | 74.2% |
| LMArena WebDev | 1362 | 1439 |
| SWE-bench Multilingual | 59% | 68.7% |
| WeirdML | 39.5% | 69.9% |
| LMArena Coding | 1454 | 1482 |
| SWE-bench Verified | — | 72.9% |
| Aider Polyglot | 74.2% | — |
| SciCode | 38.9% | — |
| GSO | — | 18.6% |
| ALE-Bench | — | 1,177 |
| AlgoTune | — | 1.83 |
Agentic & Tool Use Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 37.9 (#31), Gemini 3 Pro: 40.5 (#21)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| Terminal-Bench | 39.6% | 69.4% |
| Berkeley Function Calling Leaderboard | 56.7% | 72.5% |
| Vending-Bench 2 | 1,034 | 5,478 |
| GDPval | — | 40.3% |
| Remote Labor Index | — | 1.3% |
| TheAgentCompany | 42.9% | — |
| τ²-bench Airline | — | 80.5% |
| τ²-bench Banking | — | 18% |
| τ²-bench Retail | — | 75.9% |
| τ²-bench Telecom | — | 91% |
| DeepResearch Bench | — | 46.3% |
| BALROG | — | 58.1% |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 71% |
Reasoning Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 22.0 (#185), Gemini 3 Pro: 52.5 (#28)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| ARC-AGI-2 | 4% | 31.1% |
| Kagi LLM Benchmark | 52.2% | 80.1% |
| NYT Connections (extended) | 36.7% | 94.4% |
| ARC-AGI-1 | 57% | 75% |
| CritPt | 2.9% | 6.9% |
| Chess Puzzles | 14% | 31% |
| LMArena Hard Prompts | 1433 | 1481 |
| Epoch Capabilities Index | 146.28 | 152.96 |
| SimpleBench | — | 76.4% |
| EnigmaEval | — | 18.2% |
| Thematic Generalization | 65% | — |
| DTBench | 87.7% | — |
| LMCA | 29.1% | — |
| ForecastBench | — | 61.2 |
Math Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 40.8 (#75), Gemini 3 Pro: 48.8 (#50)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| MathArena Final-Answer Competitions | 57.7% | 67% |
| OTIS Mock AIME 2024-2025 | 87.8% | 91.4% |
| ProofBench | 8% | 20% |
| LMArena Math | 1435 | 1475 |
| FrontierMath (Feb 2025 set) | 22.1% | 37.6% |
| FrontierMath Tier 4 (v1) | 2.1% | 18.8% |
| Omni-MATH | — | 55.5% |
Knowledge Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 51.6 (#61), Gemini 3 Pro: 64.3 (#15)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| GPQA Diamond | 83.4% | 92.6% |
| Vectara Hallucination Rate | 5.3% | 13.6% |
| LMArena Expert | 1436 | 1475 |
| Humanity's Last Exam | — | 37.5% |
| MMLU-Pro | — | 90.3% |
| GPQA (HELM) | — | 80.3% |
Multimodal Not comparable
DeepSeek-V3.2-Exp: —, Gemini 3 Pro: 56.9 (#1)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| LMArena Vision | — | 1305 |
| GeoBench | — | 84% |
| VPCT | — | 91% |
| LMArena Document | — | 1434 |
Multilingual Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 52.4 (#85), Gemini 3 Pro: 57.1 (#14)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| LMArena Non-English | 1409 | 1474 |
| LMArena Chinese | 1461 | 1522 |
| LMArena French | 1435 | 1492 |
| LMArena German | 1440 | 1517 |
| LMArena Japanese | 1377 | 1513 |
| LMArena Korean | 1372 | 1449 |
| LMArena Russian | 1423 | 1493 |
| LMArena Spanish | 1441 | 1468 |
Instruction Following Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 74.7 (#89), Gemini 3 Pro: 76.4 (#42)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| LMArena Instruction Following | 1413 | 1458 |
| IFEval | — | 87.7% |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#15), Gemini 3 Pro: 44.0 (#75)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| CL-bench | 13.2% | 15.8% |
| LMArena Longer Query | 1428 | 1472 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Gemini 3 Pro leads
DeepSeek-V3.2-Exp: 62.5 (#73), Gemini 3 Pro: 66.6 (#34)
| Benchmark | DeepSeek-V3.2-Exp | Gemini 3 Pro |
|---|---|---|
| LMArena Text | 1425 | 1479 |
| LMArena Creative Writing | 1403 | 1482 |
| EQ-Bench Creative Writing | 1515 | 1525 |
| LMArena Multi-Turn | 1427 | 1484 |
| WildBench | — | 85.9% |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Gemini 3 Pro?
Gemini 3 Pro is the stronger model overall, scoring 54.7 to 44.9 on the Noometry Index.
Is DeepSeek-V3.2-Exp or Gemini 3 Pro better for coding?
Gemini 3 Pro scores higher on coding benchmarks: 52.1 versus 46.4 in the Noometry coding category.
How many benchmarks do DeepSeek-V3.2-Exp and Gemini 3 Pro share?
40 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 48 scored results on Noometry and Gemini 3 Pro has 67.