Model benchmarks
gemma4:12b local LLM performance
As of August 2026, gemma4:12b runs at up to 63.1 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
11.9B
Peak speed
63.1 tok/s
Average speed
51.9 tok/s
Min memory
7.8 GB
Max context
196.608 tokens
Avg runtime / run
3m 55s
Avg quality
70.2
Benchmark runs
3
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 3 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 68.1 | 70.2 | 72.2 |
| Agent Workflow | 71.9 | 75.3 | 78.6 |
| Code Generation | 55.4 | 58.0 | 59.8 |
| Role Play & Narrative | 78.2 | 79.7 | 80.9 |
| Research & Analysis | 62.4 | 67.9 | 74.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination gemma4:12b has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 7900 XTX | Ollama | Q4_K_M | 63.1 tok/s | 63.1 tok/s | 8.2 GB | 196.608 tokens | 72.4 | 1 |
| CPU only | Ollama | Q4_K_M | 46.3 tok/s | 46.3 tok/s | 7.8 GB | 32.768 tokens | 69.1 | 2 |
Benchmark runs
All 3 gemma4:12b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is gemma4:12b good for coding?
- In our benchmarks, gemma4:12b scores 58.0/100 for coding. It runs at about 51.9 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
- Is gemma4:12b good for agentic (tool-using) tasks?
- In our benchmarks, gemma4:12b scores 75.3/100 for agentic workflows. It runs at about 51.9 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
- How fast is gemma4:12b for local inference?
- Across 3 community benchmark runs, gemma4:12b reaches up to 63.1 tok/s and averages 51.9 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
- How much memory does gemma4:12b need?
- The leanest observed configuration used about 7.8 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run gemma4:12b?
- Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.