Model comparison
DeepSeek-V2.5 (Sep 2024) vs Gemma 3 12B
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 32.1 on the Noometry Index.
Last verified . 12 shared benchmarks.
DeepSeek
GoogleSummary
- They share 12 benchmarks with published results for both. DeepSeek-V2.5 (Sep 2024) scores higher in 4 categories and Gemma 3 12B in 4 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 35.9 to 22.3.
Side by side
| DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B | |
|---|---|---|
| Provider | DeepSeek | |
| Noometry Index | 37.6 | 32.1 |
| Released | 2024-09-06 | 2025-03-12 |
| Weights | Open | Open |
| Context window | — | 131K |
| Max output | — | 8K |
| Input $ / M tokens | — | $0.05 |
| Output $ / M tokens | — | $0.15 |
| Results tracked | 22 | 24 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V2.5 (Sep 2024): 31.7 (#281), Gemma 3 12B: 31.7 (#280)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Coding | 1309 | 1281 |
| Aider Polyglot | 17.8% | — |
| SciCode | — | 17.4% |
| BigCodeBench Instruct | 48.6% | — |
| BigCodeBench Complete | 53.2% | — |
| HumanEval+ | 83.5% | — |
| MBPP+ | 74.1% | — |
Agentic & Tool Use Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemma 3 12B: 25.5 (#108)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 30.4% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 25.6 (#145), Gemma 3 12B: 15.7 (#313)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Hard Prompts | 1289 | 1309 |
| CritPt | — | 0% |
| Chess Puzzles | — | 0% |
| DTBench | — | 48.8% |
| LMCA | — | 4.5% |
| Epoch Capabilities Index | — | 123.5 |
Math DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 35.9 (#177), Gemma 3 12B: 22.3 (#279)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Math | 1288 | 1307 |
| OTIS Mock AIME 2024-2025 | — | 16.7% |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 34.8 (#193), Gemma 3 12B: 26.5 (#257)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Expert | 1266 | 1248 |
| GPQA Diamond | — | 39.5% |
| Vectara Hallucination Rate | — | 4.4% |
Multimodal Not comparable
DeepSeek-V2.5 (Sep 2024): —, Gemma 3 12B: —
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| MindCube | — | 46.7% |
Multilingual Gemma 3 12B leads
DeepSeek-V2.5 (Sep 2024): 42.5 (#193), Gemma 3 12B: 45.7 (#165)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Non-English | 1273 | 1318 |
| LMArena German | 1258 | 1370 |
| LMArena Russian | 1289 | 1335 |
| LMArena Chinese | 1318 | — |
| LMArena French | 1289 | — |
| LMArena Japanese | 1228 | — |
| LMArena Korean | 1209 | — |
| LMArena Spanish | 1248 | — |
Instruction Following Gemma 3 12B leads
DeepSeek-V2.5 (Sep 2024): 67.5 (#194), Gemma 3 12B: 68.6 (#186)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Instruction Following | 1280 | 1299 |
Long Context Too close to call
DeepSeek-V2.5 (Sep 2024): 39.5 (#174), Gemma 3 12B: 40.0 (#162)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Longer Query | 1301 | 1317 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
DeepSeek-V2.5 (Sep 2024): 49.8 (#187), Gemma 3 12B: 47.5 (#209)
| Benchmark | DeepSeek-V2.5 (Sep 2024) | Gemma 3 12B |
|---|---|---|
| LMArena Text | 1294 | 1334 |
| LMArena Creative Writing | 1285 | 1331 |
| LMArena Multi-Turn | 1297 | 1334 |
| EQ-Bench Creative Writing | — | 1126 |
Frequently asked questions
Is DeepSeek-V2.5 (Sep 2024) better than Gemma 3 12B?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 32.1 on the Noometry Index.
Is DeepSeek-V2.5 (Sep 2024) or Gemma 3 12B better for coding?
They score almost the same on coding (31.7 vs 31.7); test both on your own repository before choosing.
How many benchmarks do DeepSeek-V2.5 (Sep 2024) and Gemma 3 12B share?
12 benchmarks have published results for both models. DeepSeek-V2.5 (Sep 2024) has 22 scored results on Noometry and Gemma 3 12B has 24.