Model comparison
DeepSeek-V2.5 (Sep 2024) vs Gemini 1.5 Flash (May 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 33.2 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeek
GoogleSummary
- They share 21 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 6 categories and Gemini 1.5 Flash (May 2024) in 2 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 22.1.
- The biggest single-benchmark swing is BigCodeBench Instruct: 48.6% for DeepSeek-V2.5 (Sep 2024) and 43.5% for Gemini 1.5 Flash (May 2024).
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 37.6 | 33.2 |
| Released | 2024-09-06 | 2024-05-14 |
| Weights | Open | Proprietary |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 22 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Gemini 1.5 Flash (May 2024) leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Gemini 1.5 Flash (May 2024): 34.4 (#236)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| BigCodeBench Instruct | 48.6% | 43.5% |
| LMArena Coding | 1309 | 1261 |
| BigCodeBench Complete | 53.2% | 55.1% |
| HumanEval+ | 83.5% | 75.6% |
| MBPP+ | 74.1% | 67.5% |
| Aider Polyglot | 17.8% | — |
| WeirdML | — | 24.9% |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemini 1.5 Flash (May 2024): 26.6 (#102)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| BALROG | — | 14.6% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Gemini 1.5 Flash (May 2024): 21.7 (#215)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1257 |
| DTBench | — | 53.8% |
| Epoch Capabilities Index | — | 129.36 |
| ForecastBench | — | 53.9 |
| PIQA | — | 87.5% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Gemini 1.5 Flash (May 2024): 22.1 (#281)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Math | 1288 | 1269 |
| OTIS Mock AIME 2024-2025 | — | 16.3% |
| Omni-MATH | — | 30.4% |
| MATH Level 5 | — | 61.9% |
| FrontierMath (Feb 2025 set) | — | 0% |
| GSM8K | — | 82.4% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Gemini 1.5 Flash (May 2024): 26.2 (#260)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Expert | 1266 | 1233 |
| GPQA Diamond | — | 47.3% |
| MMLU-Pro | — | 67.8% |
| GPQA (HELM) | — | 43.7% |
| BoolQ | — | 85.8% |
| MMLU | — | 77.9% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemini 1.5 Flash (May 2024): 36.0 (#81)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Vision | — | 1141 |
| Video-MME | — | 70.3% |
| GeoBench | — | 76% |
Multilingual Too close to call
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Gemini 1.5 Flash (May 2024): 42.9 (#189)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Non-English | 1273 | 1278 |
| LMArena Chinese | 1318 | 1295 |
| LMArena French | 1289 | 1258 |
| LMArena German | 1258 | 1262 |
| LMArena Japanese | 1228 | 1252 |
| LMArena Korean | 1209 | 1221 |
| LMArena Russian | 1289 | 1288 |
| LMArena Spanish | 1248 | 1243 |
Instruction Following Too close to call
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Gemini 1.5 Flash (May 2024): 66.8 (#205)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Instruction Following | 1280 | 1258 |
| IFEval | — | 83.1% |
Long Context Too close to call
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Gemini 1.5 Flash (May 2024): 39.0 (#187)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Longer Query | 1301 | 1284 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Gemini 1.5 Flash (May 2024): 48.7 (#196)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemini 1.5 Flash (May 2024) |
|---|---|---|
| LMArena Text | 1294 | 1287 |
| LMArena Creative Writing | 1285 | 1285 |
| LMArena Multi-Turn | 1297 | 1253 |
| WildBench | — | 79.2% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Gemini 1.5 Flash (May 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 33.2 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Gemini 1.5 Flash (May 2024) better for coding?
Gemini 1.5 Flash (May 2024) scores higher on coding benchmarks: 34.4 versus 31.7 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Gemini 1.5 Flash (May 2024) share?
21 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Gemini 1.5 Flash (May 2024) has 42.