Model benchmarks
unsloth/gemma-4-26B-A4B-it-qat-GGUF local LLM performance
As of October 2026, unsloth/gemma-4-26B-A4B-it-qat-GGUF runs at up to 57.7 tok/s for local inference (best of 4 community benchmark runs across 3 GPUs).
Model size
26B
Peak speed
57.7 tok/s
Average speed
42.6 tok/s
Avg PP
387.3 tok/s
Min memory
16.0 GB
Max context
147,072 tokens
Avg output / run
15,198 tokens
Avg runtime / run
6m 39s
Avg quality
67.0
Benchmark runs
4
GPUs tested
3
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 4 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 | 62.6 | 67.0 | 70.6 |
| Agent Workflow | 68.0 | 73.4 | 79.8 |
| Code Generation | 37.2 | 56.6 | 69.1 |
| Role Play & Narrative | 74.0 | 79.4 | 82.8 |
| Research & Analysis | 50.3 | 58.8 | 63.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/gemma-4-26B-A4B-it-qat-GGUF 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3070 Ti | Unsloth Studio | UD-Q4_K_XL | 57.7 tok/s | 57.7 tok/s | n/a | 8,192 tokens | 66.3 | 1 |
| NVIDIA GeForce RTX 3080 | Unsloth Studio | UD-Q4_K_XL | 40.8 tok/s | 36.1 tok/s | 26.0 GB | 147,072 tokens | 66.4 | 2 |
| AMD Radeon RX 9070/9070 XT/9070 GRE | llama.cpp | — | 40.4 tok/s | 40.4 tok/s | 16.0 GB | 32,768 tokens | 69.1 | 1 |
Benchmark runs
All 4 unsloth/gemma-4-26B-A4B-it-qat-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/gemma-4-26B-A4B-it-qat-GGUF good for coding?
- In our benchmarks, unsloth/gemma-4-26B-A4B-it-qat-GGUF scores 56.6/100 for coding. It runs at about 42.6 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~75.9 tok/s) and still scores well for coding (82.9/100).
- Is unsloth/gemma-4-26B-A4B-it-qat-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/gemma-4-26B-A4B-it-qat-GGUF scores 73.4/100 for agentic workflows. It runs at about 42.6 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 unsloth/gemma-4-26B-A4B-it-qat-GGUF for local inference?
- Across 4 community benchmark runs, unsloth/gemma-4-26B-A4B-it-qat-GGUF reaches up to 57.7 tok/s and averages 42.6 tok/s, with the fastest results on NVIDIA GeForce RTX 3070 Ti.
- How much memory does unsloth/gemma-4-26B-A4B-it-qat-GGUF need?
- The leanest observed configuration used about 16.0 GB of memory (quantizations tested: UD-Q4_K_XL).
- Which tools have been used to run unsloth/gemma-4-26B-A4B-it-qat-GGUF?
- Benchmarks were submitted using Unsloth Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.