Model benchmarks

gemma4:26b local LLM performance

As of June 2026, gemma4:26b runs at up to 50.6 tok/s for local inference (best of 21 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

25.8B

Peak speed

50.6 tok/s

Average speed

28.6 tok/s

Min memory

16.2 GB

Max context

10,432 tokens

Best quality

94.6

Benchmark runs

21

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination gemma4:26b has been benchmarked on, ranked by peak token generation speed. Last updated June 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 7900 XTXOllamaQ4_K_M50.6 tok/s28.6 tok/s16.2 GB10,432 tokens61.621

Frequently asked questions

How fast is gemma4:26b for local inference?
Across 21 community benchmark runs, gemma4:26b reaches up to 50.6 tok/s and averages 28.6 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does gemma4:26b need?
The leanest observed configuration used about 16.2 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run gemma4:26b?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.