Model benchmarks
unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL local LLM performance
As of October 2026, unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL runs at up to 52.7 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
31B
Peak speed
52.7 tok/s
Average speed
49.6 tok/s
Avg PP
360.5 tok/s
Min memory
1100.4 GB
Max context
65,536 tokens
Avg output / run
9,760 tokens
Avg runtime / run
3m 27s
Avg quality
73.8
Benchmark runs
2
GPUs tested
2
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 2 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 | 72.2 | 73.8 | 75.3 |
| Agent Workflow | 80.3 | 82.6 | 84.9 |
| Code Generation | 63.1 | 67.6 | 72.0 |
| Role Play & Narrative | 62.6 | 67.5 | 72.5 |
| Research & Analysis | 74.1 | 77.4 | 80.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Advanced Micro Devices, Inc. [AMD/ATI] Device 7551 | llama.cpp | Q4_K | 52.7 tok/s | 52.7 tok/s | 1100.4 GB | 65,536 tokens | 75.5 | 1 |
| NVIDIA GeForce RTX 3080 Ti | llama.cpp | Q4_K | 46.4 tok/s | 46.4 tok/s | n/a | 65,536 tokens | 72.1 | 1 |
Benchmark runs
All 2 unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL good for coding?
- In our benchmarks, unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL scores 67.6/100 for coding. It runs at about 49.6 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL scores 82.6/100 for agentic workflows. It runs at about 49.6 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 unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL for local inference?
- Across 2 community benchmark runs, unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL reaches up to 52.7 tok/s and averages 49.6 tok/s, with the fastest results on Advanced Micro Devices, Inc. [AMD/ATI] Device 7551.
- How much memory does unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL need?
- The leanest observed configuration used about 1100.4 GB of memory (quantizations tested: Q4_K).
- Which tools have been used to run unsloth/gemma-4-31B-it-qat-GGUF:UD-Q4_K_XL?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.