gemma4:26b

gemma4 · 25.8B · Q4_K_M

THINKING MODEL

Mac mini (Apple M4)

24 GB · macOS 26.3

Tested on April 3, 2026 · Submitted by cryptoepops

Top 95% Compare

Global Score

33 /100

Not Rec.

Hardware Fit

50/100

Quality

25/100

Get this model

🦙

Ollama

ollama pull gemma4:26b

View on Ollama Library

ollama.com/library/gemma4

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: Mac mini
CPU: Apple M4
Cores: 10 total (4 perf + 6 eff)
Frequency: 2.4 GHz
RAM: 24 GB LPDDR5
GPU: Apple M4
OS: macOS 26.3
Arch: arm64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 20.0
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±1.8
First chunk latency Delay before the first streamed chunk arrives from runtime: 941 ms
Time to first token How fast the model starts responding — lower is better: 30.0 s
Load time Time to load the model into memory before inference: 0.2 s
Memory usage RAM consumed during inference — percentage of total system memory: 20.3 GB (85%)
Total tokens Total number of tokens generated across all benchmark prompts: 1455
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~736

Score breakdown

Speed

42/50

Time to first token

0/20

Memory

8/30

Quality

Reasoning

14/20

Coding

1/20

Instruction following

6/20

Structured output

3/15

Math

0/15

Multilingual

1/10

Category levels

Reasoning: Adequate Coding: Poor Instruction Following: Weak Structured Output: Poor Math: Poor Multilingual: Poor

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.20.0
Model format: GGUF
Hardware profile: ENTRY
Result hash: e168e5a1a05941783d10eee10b5e9ca6cea4f1043d51abb52ba55dfa2ca60bd2

Interpretation

Hardware fit: 50/100. Overall suitability: NOT RECOMMENDED (Global 33/100). Category profile: Reasoning: Adequate, Coding: Poor, Instruction Following: Weak, Structured Output: Poor, Math: Poor, Multilingual: Poor.

Warnings

Significant swap activity during benchmark (+0.6 GB). Model may exceed available RAM — results are severely degraded.

Disqualifiers

Time to first token too high: 30000ms (maximum: 21386ms for ENTRY profile)

Bench Environment

Power: AC Swap delta: +0.6 GB CPU load: avg 49% (peak 54%)

Run yours now

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest