qwen3:8b

qwen3 · 8.2B · Q4_K_M

THINKING MODEL

MacBook Air (Apple M4)

32 GB · macOS 26.3

Tested on March 2, 2026

Top 86% Compare

Global Score

50 /100

Marginal

Hardware Fit

67/100

Quality

38/100

Get this model

🦙

Ollama

ollama pull qwen3:8b

View on Ollama Library

ollama.com/library/qwen3

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: MacBook Air
CPU: Apple M4
Cores: 10 total (4 perf + 6 eff)
Frequency: 2.4 GHz
RAM: 32 GB LPDDR5
GPU: Apple M4
OS: macOS 26.3
Arch: arm64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 18.7
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±0.5
Time to first token How fast the model starts responding — lower is better: 12.1 s
Load time Time to load the model into memory before inference: 1.2 s
Memory usage RAM consumed during inference — percentage of total system memory: 9.5 GB (30%)
Total tokens Total number of tokens generated across all benchmark prompts: 1429
Thinking tokens (est.) Estimated internal reasoning tokens — models with "thinking" generate more tokens: ~968

Score breakdown

Speed

31/40

Time to first token

6/30

Memory

30/30

Quality

Reasoning

5/20

Coding

5/20

Instruction following

4/20

Structured output

12/15

Math

2/15

Multilingual

10/10

Category levels

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

Metadata

Spec version: 0.2.0
Runtime: Ollama 0.17.4
Model format: GGUF
Hardware profile: BALANCED
Result hash: 60abe32ab0495a1c48414134fa3f90cbc8d48dfd8440c020ff32b2866f19f4b5

Interpretation

Hardware fit: 67/100. Overall suitability: MARGINAL (Global 50/100). Category profile: Reasoning: Weak, Coding: Weak, Instruction Following: Poor, Structured Output: Strong, Math: Poor, Multilingual: Strong.

Run yours now

npx metrillm@latest bench

Requires Node 20+ and Ollama running