Is gemma4:e4b good for coding?

In our benchmarks, gemma4:e4b scores 46.9/100 for coding. It runs at about 71.4 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:e4b good for agentic (tool-using) tasks?

In our benchmarks, gemma4:e4b scores 76.1/100 for agentic workflows. It runs at about 71.4 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:e4b for local inference?

Across 10 community benchmark runs, gemma4:e4b reaches up to 115.4 tok/s and averages 71.4 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.

How much memory does gemma4:e4b need?

The leanest observed configuration used about 3.0 GB of memory (quantizations tested: Q4_K_M).

Which tools have been used to run gemma4:e4b?

Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.

Model benchmarks

gemma4:e4b local LLM performance

As of October 2026, gemma4:e4b runs at up to 115.4 tok/s for local inference (best of 10 community benchmark runs across 7 GPUs).

OllamaQ4_K_M

ShareReddit X

Model size

8.0B

Peak speed

115.4 tok/s

Average speed

71.4 tok/s

Avg PP

1710.1 tok/s

Min memory

3.0 GB

Max context

256,000 tokens

Avg output / run

10,031 tokens

Avg runtime / run

2m 57s

Avg quality

65.5

Benchmark runs

GPUs tested

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 10 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	58.3	65.5	72.3
Agent Workflow	62.3	76.1	87.6
Code Generation	39.4	46.9	56.5
Role Play & Narrative	61.9	76.1	87.5
Research & Analysis	52.2	62.7	72.1

Performance by hardware and tool

Every hardware/tool/quantization combination gemma4:e4b 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
AMD Radeon RX 7900 XTX	Ollama	Q4_K_M	115.4 tok/s	115.4 tok/s	10.9 GB	65,536 tokens	58.3	1
NVIDIA GeForce RTX 4070 Ti	Ollama	Q4_K_M	93.9 tok/s	93.9 tok/s	7.1 GB	65,536 tokens	70.3	1
NVIDIA GeForce RTX 3060 Ti	Ollama	Q4_K_M	92.6 tok/s	78.1 tok/s	3.0 GB	256,000 tokens	71.5	2
CPU only	Ollama	Q4_K_M	75.2 tok/s	75.2 tok/s	3.2 GB	8,192 tokens	61.8	1
NVIDIA GeForce RTX 4070 Laptop GPU	Ollama	Q4_K_M	57.6 tok/s	57.6 tok/s	3.1 GB	8,192 tokens	64.5	1
Apple M4 Pro	Ollama	Q4_K_M	55.4 tok/s	54.8 tok/s	3.1 GB	131,072 tokens	60.6	2
AMD Radeon RX 9060 XT	Ollama	Q4_K_M	54.0 tok/s	54.0 tok/s	3.3 GB	65,536 tokens	67.5	1
NVIDIA GeForce RTX 4060 Laptop GPU	Ollama	Q4_K_M	51.8 tok/s	51.8 tok/s	3.0 GB	8,192 tokens	68.3	1

Benchmark runs

All 10 gemma4:e4b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is gemma4:e4b good for coding?: In our benchmarks, gemma4:e4b scores 46.9/100 for coding. It runs at about 71.4 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:e4b good for agentic (tool-using) tasks?: In our benchmarks, gemma4:e4b scores 76.1/100 for agentic workflows. It runs at about 71.4 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:e4b for local inference?: Across 10 community benchmark runs, gemma4:e4b reaches up to 115.4 tok/s and averages 71.4 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does gemma4:e4b need?: The leanest observed configuration used about 3.0 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run gemma4:e4b?: Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.