Model benchmarks

hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S local LLM performance

As of November 2026, hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S runs at up to 71.4 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).

OllamaIQ3_S
ShareRedditX

Model size

25.2B

Peak speed

71.4 tok/s

Average speed

71.2 tok/s

Avg PP

1231.8 tok/s

Min memory

10.9 GB

Max context

32,768 tokens

Avg output / run

16,478 tokens

Avg runtime / run

4m 1s

Avg quality

71.5

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.

TaskP5 (low)AvgP95 (high)
Overall67.071.576.2
Agent Workflow74.876.077.8
Code Generation63.169.473.6
Role Play & Narrative72.773.574.4
Research & Analysis55.767.080.7

Performance by hardware and tool

Every hardware/tool/quantization combination hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
Advanced Micro Devices, Inc. [AMD/ATI] Device 7590OllamaIQ3_S71.4 tok/s71.2 tok/s10.9 GB32,768 tokens71.53

Benchmark runs

All 3 hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S good for coding?
In our benchmarks, hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S scores 69.4/100 for coding. It runs at about 71.2 tok/s, so if you want more speed, qwen3.8-27b-Swift1.5-Uncensored-ATX-AI3 is faster (~85.5 tok/s) and still scores well for coding (84.8/100).
Is hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S good for agentic (tool-using) tasks?
In our benchmarks, hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S scores 76.0/100 for agentic workflows. It runs at about 71.2 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 hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S for local inference?
Across 3 community benchmark runs, hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S reaches up to 71.4 tok/s and averages 71.2 tok/s, with the fastest results on Advanced Micro Devices, Inc. [AMD/ATI] Device 7590.
How much memory does hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S need?
The leanest observed configuration used about 10.9 GB of memory (quantizations tested: IQ3_S).
Which tools have been used to run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-IQ3_S?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.