Model comparison
DeepSeek-V2.5 (Sep 2024) vs Gemma 3 27B
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.8 on the Noometry Index.
Last verified . 18 shared benchmarks.
DeepSeek
GoogleSummary
- They share 18 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 5 categories and Gemma 3 27B in 3 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in long context, where DeepSeek-V2.5 (Sep 2024) leads 39.5 to 27.6.
- The biggest single-benchmark swing is Aider Polyglot: 17.8% for DeepSeek-V2.5 (Sep 2024) and 4.9% for Gemma 3 27B.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 37.6 | 30.8 |
| Released | 2024-09-06 | 2025-03-11 |
| Weights | Open | Open |
| Context window | — | 131K |
| Max output | — | 8K |
| Input $ / M tokens | — | $0.08 |
| Output $ / M tokens | — | $0.16 |
| Results tracked | 22 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Gemma 3 27B: 22.5 (#334)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| Aider Polyglot | 17.8% | 4.9% |
| LMArena Coding | 1309 | 1322 |
| SciCode | — | 21.2% |
| BigCodeBench Instruct | 48.6% | — |
| LiveBench Coding | — | 39.9% |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemma 3 27B: 25.1 (#110)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 29.5% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Gemma 3 27B: 16.7 (#301)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1340 |
| Kagi LLM Benchmark | — | 40.4% |
| CritPt | — | 0% |
| Chess Puzzles | — | 0% |
| LiveBench Reasoning | — | 43.8% |
| DTBench | — | 52.5% |
| LiveBench Data Analysis | — | 51.5% |
| LMCA | — | 12.3% |
| Epoch Capabilities Index | — | 130.04 |
| LiveBench | — | 50% |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Gemma 3 27B: 25.9 (#265)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Math | 1288 | 1312 |
| OTIS Mock AIME 2024-2025 | — | 22.5% |
| LiveBench Math | — | 55.4% |
| MATH Level 5 | — | 74% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Gemma 3 27B: 25.5 (#261)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Expert | 1266 | 1304 |
| GPQA Diamond | — | 47.7% |
| Confabulations | — | 40.3% |
| Vectara Hallucination Rate | — | 7.4% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemma 3 27B: 32.6 (#100)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Vision | — | 1164 |
| GeoBench | — | 52% |
Multilingual Gemma 3 27B leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Gemma 3 27B: 46.9 (#155)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Non-English | 1273 | 1334 |
| LMArena Chinese | 1318 | 1346 |
| LMArena French | 1289 | 1368 |
| LMArena German | 1258 | 1362 |
| LMArena Japanese | 1228 | 1287 |
| LMArena Korean | 1209 | 1308 |
| LMArena Russian | 1289 | 1349 |
| LMArena Spanish | 1248 | 1349 |
Instruction Following Gemma 3 27B leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Gemma 3 27B: 70.6 (#160)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1321 |
| LiveBench Instruction Following | — | 74.9% |
Long Context DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Gemma 3 27B: 27.6 (#293)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Longer Query | 1301 | 1333 |
| Fiction.LiveBench | — | 33.3% |
Writing & Preference Gemma 3 27B leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Gemma 3 27B: 52.5 (#168)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 27B |
|---|---|---|
| LMArena Text | 1294 | 1358 |
| LMArena Creative Writing | 1285 | 1346 |
| LMArena Multi-Turn | 1297 | 1345 |
| Short-Story Creative Writing | — | 79.9% |
| EQ-Bench Creative Writing | — | 1266 |
| LiveBench Language | — | 34.6% |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Gemma 3 27B?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.8 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Gemma 3 27B better for coding?
DeepSeek-V2.5 (Sep 2024) scores higher on coding benchmarks: 31.7 versus 22.5 in the Noometry coding category.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Gemma 3 27B share?
18 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Gemma 3 27B has 43.