Model comparison
Gemma 3 12B vs Qwen3.5 397B-A17B
Qwen3.5 397B-A17B is the stronger model overall, scoring 46.0 to 32.1 on the Noometry Index. Gemma 3 12B costs 18× less per token, which makes it the better buy when Qwen3.5 397B-A17B's lead doesn't matter for your workload.
Last verified . 19 shared benchmarks.
Google
Alibaba (Qwen)Summary
- They share 19 benchmarks with published results for both. Gemma 3 12B scores higher in 0 categories and Qwen3.5 397B-A17B in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen3.5 397B-A17B leads 53.3 to 26.5.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 16.7% for Gemma 3 12B and 88.9% for Qwen3.5 397B-A17B.
- Gemma 3 12B is cheaper at $0.05 / $0.15 per million input/output tokens, against $0.60 / $3.60 for Qwen3.5 397B-A17B.
- Qwen3.5 397B-A17B accepts more context: 262K tokens versus 131K.
Side by side
| Gemma 3 12B | Qwen3.5 397B-A17B | |
|---|---|---|
| Provider | Alibaba (Qwen) | |
| Noometry Index | 32.1 | 46.0 |
| Released | 2025-03-12 | 2026-02-01 |
| Weights | Open | Open |
| Context window | 131K | 262K |
| Max output | 8K | 66K |
| Input $ / M tokens | $0.05 | $0.60 |
| Output $ / M tokens | $0.15 | $3.60 |
| Results tracked | 24 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.5 397B-A17B leads
Gemma 3 12B: 31.7 (#280), Qwen3.5 397B-A17B: 42.0 (#114)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Coding | 1281 | 1465 |
| LMArena WebDev | — | 1400 |
| SciCode | 17.4% | — |
Agentic & Tool Use Qwen3.5 397B-A17B leads
Gemma 3 12B: 25.5 (#108), Qwen3.5 397B-A17B: 33.3 (#53)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| APEX-Agents | — | 24.9% |
| Berkeley Function Calling Leaderboard | 30.4% | — |
| τ²-bench Airline | — | 81.5% |
| τ²-bench Banking | — | 9.8% |
| τ²-bench Retail | — | 84.4% |
| τ²-bench Telecom | — | 97.8% |
Reasoning Qwen3.5 397B-A17B leads
Gemma 3 12B: 15.7 (#313), Qwen3.5 397B-A17B: 34.5 (#70)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| Chess Puzzles | 0% | 13% |
| LMArena Hard Prompts | 1309 | 1448 |
| DTBench | 48.8% | 87.5% |
| LMCA | 4.5% | 37.9% |
| Epoch Capabilities Index | 123.5 | 146.65 |
| Kagi LLM Benchmark | — | 73.7% |
| NYT Connections (extended) | — | 58.9% |
| CritPt | 0% | — |
| Thematic Generalization | — | 65.1% |
| Mystery Game Puzzles | — | 18% |
Math Qwen3.5 397B-A17B leads
Gemma 3 12B: 22.3 (#279), Qwen3.5 397B-A17B: 46.1 (#73)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 16.7% | 88.9% |
| LMArena Math | 1307 | 1454 |
| FrontierMath (Tiers 1-3) | — | 31.2% |
Knowledge Qwen3.5 397B-A17B leads
Gemma 3 12B: 26.5 (#257), Qwen3.5 397B-A17B: 53.3 (#58)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| GPQA Diamond | 39.5% | 86.4% |
| LMArena Expert | 1248 | 1462 |
| Vectara Hallucination Rate | 4.4% | — |
Multimodal Not comparable
Gemma 3 12B: —, Qwen3.5 397B-A17B: 40.7 (#44)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Vision | — | 1263 |
| MindCube | 46.7% | — |
Multilingual Qwen3.5 397B-A17B leads
Gemma 3 12B: 45.7 (#165), Qwen3.5 397B-A17B: 53.7 (#59)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Non-English | 1318 | 1430 |
| LMArena German | 1370 | 1447 |
| LMArena Russian | 1335 | 1429 |
| LMArena Chinese | — | 1500 |
| LMArena French | — | 1461 |
| LMArena Japanese | — | 1426 |
| LMArena Korean | — | 1384 |
| LMArena Spanish | — | 1441 |
Instruction Following Qwen3.5 397B-A17B leads
Gemma 3 12B: 68.6 (#186), Qwen3.5 397B-A17B: 75.0 (#77)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Instruction Following | 1299 | 1424 |
Long Context Qwen3.5 397B-A17B leads
Gemma 3 12B: 40.0 (#162), Qwen3.5 397B-A17B: 44.1 (#74)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Longer Query | 1317 | 1442 |
Writing & Preference Qwen3.5 397B-A17B leads
Gemma 3 12B: 47.5 (#209), Qwen3.5 397B-A17B: 62.3 (#79)
| Benchmark | Gemma 3 12B | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Text | 1334 | 1438 |
| LMArena Creative Writing | 1331 | 1401 |
| EQ-Bench Creative Writing | 1126 | 1478 |
| LMArena Multi-Turn | 1334 | 1446 |
Frequently asked questions
Is Gemma 3 12B better than Qwen3.5 397B-A17B?
Qwen3.5 397B-A17B is the stronger model overall, scoring 46.0 to 32.1 on the Noometry Index. Gemma 3 12B costs 18× less per token, which makes it the better buy when Qwen3.5 397B-A17B's lead doesn't matter for your workload.
Which is cheaper, Gemma 3 12B or Qwen3.5 397B-A17B?
Gemma 3 12B is cheaper. It lists at $0.05 per million input tokens and $0.15 per million output tokens; Qwen3.5 397B-A17B lists at $0.60 and $3.60.
Is Gemma 3 12B or Qwen3.5 397B-A17B better for coding?
Qwen3.5 397B-A17B scores higher on coding benchmarks: 42.0 versus 31.7 in the Noometry coding category.
Which has the bigger context window?
Qwen3.5 397B-A17B does, with 262K tokens against 131K.
How many benchmarks do Gemma 3 12B and Qwen3.5 397B-A17B share?
19 benchmarks have published results for both models. Gemma 3 12B has 24 scored results on Noometry and Qwen3.5 397B-A17B has 36.