Model benchmarks
unsloth/gemma-4-E4B-it-qat-mobile-GGUF local LLM performance
As of September 2026, unsloth/gemma-4-E4B-it-qat-mobile-GGUF runs at up to 77.4 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
4B
Peak speed
77.4 tok/s
Average speed
74.6 tok/s
Avg PP
1102.6 tok/s
Min memory
11.2 GB
Max context
262,144 tokens
Avg output / run
7,639 tokens
Avg runtime / run
1m 47s
Avg quality
27.4
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 | 24.0 | 27.4 | 29.5 |
| Agent Workflow | 24.6 | 32.4 | 39.9 |
| Code Generation | 4.5 | 11.2 | 16.1 |
| Role Play & Narrative | 38.4 | 41.4 | 43.4 |
| Research & Analysis | 23.2 | 24.6 | 26.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/gemma-4-E4B-it-qat-mobile-GGUF 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 4070 | Unsloth Studio | UD-Q2_K_XL | 77.4 tok/s | 77.4 tok/s | 11.2 GB | 262,144 tokens | 29.3 | 1 |
| NVIDIA GeForce RTX 4070 | llama.cpp | — | 73.6 tok/s | 73.1 tok/s | n/a | 131,072 tokens | 26.4 | 2 |
Benchmark runs
All 3 unsloth/gemma-4-E4B-it-qat-mobile-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/gemma-4-E4B-it-qat-mobile-GGUF good for coding?
- In our benchmarks, unsloth/gemma-4-E4B-it-qat-mobile-GGUF scores 11.2/100 for coding. It runs at about 74.6 tok/s, so if you want more speed, Ornith-1.5-35B-A3B.APEX-I-NanoPlus is faster (~110.5 tok/s) and still scores well for coding (81.1/100).
- Is unsloth/gemma-4-E4B-it-qat-mobile-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/gemma-4-E4B-it-qat-mobile-GGUF scores 32.4/100 for agentic workflows. It runs at about 74.6 tok/s, so if you want more speed, swift-iq2_xs is faster (~91.5 tok/s) and still scores well for agentic workflows (91.4/100).
- How fast is unsloth/gemma-4-E4B-it-qat-mobile-GGUF for local inference?
- Across 3 community benchmark runs, unsloth/gemma-4-E4B-it-qat-mobile-GGUF reaches up to 77.4 tok/s and averages 74.6 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
- How much memory does unsloth/gemma-4-E4B-it-qat-mobile-GGUF need?
- The leanest observed configuration used about 11.2 GB of memory (quantizations tested: UD-Q2_K_XL).
- Which tools have been used to run unsloth/gemma-4-E4B-it-qat-mobile-GGUF?
- Benchmarks were submitted using Unsloth Studio, llama.cpp. Results are community-contributed and updated as new runs arrive.