Model comparison
DeepSeek V4 Pro vs Gemini 2.0 Flash (Feb 2025)
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 35.1 on the Noometry Index.
Last verified . 25 shared benchmarks.
DeepSeek
GoogleSummary
- They share 25 benchmarks with published results for both. DeepSeek V4 Pro scores higher in 9 categories and Gemini 2.0 Flash (Feb 2025) in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek V4 Pro leads 56.5 to 15.2.
- The biggest single-benchmark swing is ARC-AGI-2: 61.3% for DeepSeek V4 Pro and 1.3% for Gemini 2.0 Flash (Feb 2025).
- DeepSeek V4 Pro has downloadable open weights; the other is API-only.
Side by side
| DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 54.3 | 35.1 |
| Released | 2026-04-24 | 2024-12-06 |
| Weights | Open | Proprietary |
| Context window | 1M | — |
| Max output | 393K | — |
| Input $ / M tokens | $0.66 | — |
| Output $ / M tokens | $1.98 | — |
| Results tracked | 48 | 54 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Pro leads
DeepSeek V4 Pro: 52.4 (#34), Gemini 2.0 Flash (Feb 2025): 28.4 (#315)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| WeirdML | 66.2% | 25.8% |
| LMArena Coding | 1470 | 1350 |
| SWE-bench Verified | 77.6% | — |
| FrontierCode | 28.6% | — |
| SWE-bench Verified (bash only) | — | 13.5% |
| Aider Polyglot | — | 38.2% |
| LMArena WebDev | 1582 | — |
| SciCode | 51% | — |
| BigCodeBench Instruct | — | 45.9% |
| LiveBench Coding | — | 63.4% |
| BigCodeBench Complete | — | 59.9% |
| CadEval | — | 30% |
| ALE-Bench | 1,403 | — |
Agentic & Tool Use DeepSeek V4 Pro leads
DeepSeek V4 Pro: 32.8 (#58), Gemini 2.0 Flash (Feb 2025): 28.1 (#92)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| APEX-Agents | 47.3% | — |
| TheAgentCompany | — | 11.4% |
| Vending-Bench 2 | 3,285 | — |
Reasoning DeepSeek V4 Pro leads
DeepSeek V4 Pro: 56.5 (#24), Gemini 2.0 Flash (Feb 2025): 15.2 (#318)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| ARC-AGI-2 | 61.3% | 1.3% |
| Kagi LLM Benchmark | 53.5% | 37.8% |
| LMArena Hard Prompts | 1461 | 1346 |
| DTBench | 93.9% | 63.2% |
| Epoch Capabilities Index | 155.31 | 135.36 |
| SimpleBench | — | 31.1% |
| NYT Connections (extended) | 91.3% | — |
| ARC-AGI-1 | 90.5% | — |
| CritPt | 18% | — |
| Chess Puzzles | 47% | — |
| EnigmaEval | — | 1.1% |
| LiveBench Reasoning | — | 78.2% |
| Mystery Game Puzzles | 43% | — |
| LiveBench Data Analysis | — | 69.4% |
| LMCA | 45.5% | — |
| Surface Evolver Bench | 40% | — |
| ForecastBench | 56.1 | — |
| LiveBench | — | 66.9% |
Math DeepSeek V4 Pro leads
DeepSeek V4 Pro: 64.8 (#30), Gemini 2.0 Flash (Feb 2025): 37.9 (#146)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.6% | 57.8% |
| LMArena Math | 1455 | 1352 |
| FrontierMath (Tiers 1-3) | 64.6% | — |
| FrontierMath Tier 4 | 26.8% | — |
| MathArena Final-Answer Competitions | 76.6% | — |
| ProofBench | 50% | — |
| Omni-MATH | — | 45.9% |
| LiveBench Math | — | 75.8% |
| MATH Level 5 | — | 82.2% |
| FrontierMath (Feb 2025 set) | — | 1.7% |
Knowledge DeepSeek V4 Pro leads
DeepSeek V4 Pro: 59.5 (#31), Gemini 2.0 Flash (Feb 2025): 32.0 (#213)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| GPQA Diamond | 91.7% | 64.1% |
| LMArena Expert | 1464 | 1339 |
| Humanity's Last Exam | — | 6.6% |
| SimpleQA Verified | 52.9% | — |
| MMLU-Pro | — | 73.7% |
| Confabulations | — | 12.4% |
| Vectara Hallucination Rate | 8.6% | — |
| GPQA (HELM) | — | 55.6% |
| MMLU | — | 79.7% |
Multimodal Not comparable
DeepSeek V4 Pro: —, Gemini 2.0 Flash (Feb 2025): 36.5 (#79)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| LMArena Vision | — | 1158 |
| GeoBench | — | 77% |
Multilingual DeepSeek V4 Pro leads
DeepSeek V4 Pro: 54.4 (#45), Gemini 2.0 Flash (Feb 2025): 47.4 (#149)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| LMArena Non-English | 1439 | 1342 |
| LMArena Chinese | 1486 | 1373 |
| LMArena French | 1472 | 1391 |
| LMArena German | 1458 | 1353 |
| LMArena Japanese | 1445 | 1294 |
| LMArena Korean | 1447 | 1313 |
| LMArena Russian | 1453 | 1351 |
| LMArena Spanish | 1458 | 1363 |
Instruction Following DeepSeek V4 Pro leads
DeepSeek V4 Pro: 76.1 (#47), Gemini 2.0 Flash (Feb 2025): 74.4 (#97)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| LMArena Instruction Following | 1448 | 1336 |
| LiveBench Instruction Following | — | 85.8% |
| IFEval | — | 84.1% |
Long Context DeepSeek V4 Pro leads
DeepSeek V4 Pro: 45.0 (#51), Gemini 2.0 Flash (Feb 2025): 38.1 (#203)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| LMArena Longer Query | 1458 | 1344 |
| Fiction.LiveBench | — | 61.1% |
| CL-bench Life | 13.5% | — |
Writing & Preference DeepSeek V4 Pro leads
DeepSeek V4 Pro: 65.5 (#46), Gemini 2.0 Flash (Feb 2025): 49.5 (#190)
| Benchmark | DeepSeek V4 Pro | Gemini 2.0 Flash (Feb 2025) |
|---|---|---|
| LMArena Text | 1451 | 1354 |
| LMArena Creative Writing | 1446 | 1340 |
| EQ-Bench Creative Writing | 1553 | 1128 |
| LMArena Multi-Turn | 1467 | 1350 |
| Short-Story Creative Writing | — | 73.8% |
| WildBench | — | 80% |
| EQ-Bench 4 | 1166 | — |
| LiveBench Language | — | 51.3% |
Frequently asked questions
Is DeepSeek V4 Pro better than Gemini 2.0 Flash (Feb 2025)?
DeepSeek V4 Pro is the stronger model overall, scoring 54.3 to 35.1 on the Noometry Index.
Is DeepSeek V4 Pro or Gemini 2.0 Flash (Feb 2025) better for coding?
DeepSeek V4 Pro scores higher on coding benchmarks: 52.4 versus 28.4 in the Noometry coding category.
How many benchmarks do DeepSeek V4 Pro and Gemini 2.0 Flash (Feb 2025) share?
25 benchmarks have published results for both models. DeepSeek V4 Pro has 48 scored results on Noometry and Gemini 2.0 Flash (Feb 2025) has 54.