deepseek-coder:6.7b

llama · 7B · Q4_0

Marginal

Mar 2, 2026 · Apple M4

gemma3:12b

gemma3 · 12.2B · Q4_K_M

Excellent

Mar 2, 2026 · Apple M4

Global Score

57 vs 80

Hardware Fit

98 vs 83

Quality Score

39 vs 78

Hardware

deepseek-coder:6.7b gemma3:12b

MachineMacBook AirMacBook Air

CPUApple M4Apple M4

Cores1010

RAM32 GB32 GB

GPUApple M4Apple M4

OSmacOS 26.3macOS 26.3

Archarm64arm64

Power Modebalancedbalanced

Performance

deepseek-coder:6.7b gemma3:12b

Tokens/sec24.112.7

TTFT251 ms560 ms

Load time0.9 s2.8 s

Memory usage12.6 GB11.0 GB

Memory %39%34%

HW Fit Score Breakdown

deepseek-coder:6.7b

Speed

38/40

TTFT

30/30

Memory

30/30

gemma3:12b

Speed

23/40

TTFT

30/30

Memory

30/30

Quality

deepseek-coder:6.7b

Reasoning

4/20

Coding

14/20

Instruction

9/20

Structured

6/15

Math

1/15

Multilingual

5/10

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

gemma3:12b

Reasoning

14/20

Coding

19/20

Instruction

13/20

Structured

15/15

Math

7/15

Multilingual

10/10

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

View deepseek-coder:6.7b details | View gemma3:12b details

Run yours and compare

npx metrillm@latest bench

Requires Node 20+ and Ollama or LM Studio running