Model benchmarks
gemma4:latest local LLM performance
As of September 2026, gemma4:latest runs at up to 74.4 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
8.0B
Peak speed
74.4 tok/s
Average speed
72.2 tok/s
Min memory
3.2 GB
Max context
8,192 tokens
Avg output / run
10,168 tokens
Avg runtime / run
2m 33s
Avg quality
58.3
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 | 55.1 | 58.3 | 61.8 |
| Agent Workflow | 58.6 | 64.6 | 69.5 |
| Code Generation | 33.9 | 41.5 | 50.4 |
| Role Play & Narrative | 59.3 | 69.2 | 81.9 |
| Research & Analysis | 55.4 | 58.0 | 62.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination gemma4:latest has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| CPU only | Ollama | Q4_K_M | 74.4 tok/s | 74.3 tok/s | 3.2 GB | 8,192 tokens | 58.5 | 2 |
| AMD Radeon RX 9070/9070 XT/9070 GRE | Ollama | Q4_K_M | 68.1 tok/s | 68.1 tok/s | 3.2 GB | 8,192 tokens | 58.0 | 1 |
Benchmark runs
All 3 gemma4:latest runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is gemma4:latest good for coding?
- In our benchmarks, gemma4:latest scores 41.5/100 for coding. It runs at about 72.2 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).
- Is gemma4:latest good for agentic (tool-using) tasks?
- In our benchmarks, gemma4:latest scores 64.6/100 for agentic workflows. It runs at about 72.2 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is gemma4:latest for local inference?
- Across 3 community benchmark runs, gemma4:latest reaches up to 74.4 tok/s and averages 72.2 tok/s.
- How much memory does gemma4:latest need?
- The leanest observed configuration used about 3.2 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run gemma4:latest?
- Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.