Back to leaderboard

google/gemma-3-12b

gemma3 · 12B · 4bit

Good

Mar 3, 2026 · Apple M4

granite3.1-dense:2b

granite · 2.5B · Q4_K_M

Good

Mar 2, 2026 · Apple M4

Global Score
77 vs 68
Hardware Fit
72 vs 100
Quality Score
79 vs 54

Hardware

google/gemma-3-12b granite3.1-dense:2b
MachineMacBook AirMacBook Air
CPUApple M4Apple M4
Cores1010
RAM32 GB32 GB
GPUApple M4Apple M4
OSmacOS 26.3macOS 26.3
Archarm64arm64
Power Modebalancedbalanced

Performance

google/gemma-3-12b granite3.1-dense:2b
Tokens/sec7.954.6
TTFT1.8 s121 ms
Load time0.0 s0.4 s
Memory usage0.0 GB5.6 GB
Memory %0%18%

HW Fit Score Breakdown

google/gemma-3-12b

Speed
14/40
TTFT
28/30
Memory
30/30

granite3.1-dense:2b

Speed
40/40
TTFT
30/30
Memory
30/30

Quality

google/gemma-3-12b

Reasoning
13/20
Coding
19/20
Instruction
14/20
Structured
15/15
Math
8/15
Multilingual
10/10
Reasoning: Adequate Coding: Strong Instruction Following: Adequate Structured Output: Strong Math: Adequate Multilingual: Strong

granite3.1-dense:2b

Reasoning
8/20
Coding
9/20
Instruction
13/20
Structured
14/15
Math
3/15
Multilingual
7/10
Reasoning: Weak Coding: Weak Instruction Following: Adequate Structured Output: Strong Math: Poor Multilingual: Adequate

Run yours and compare

npx metrillm@latest bench

Requires Node 20+ and Ollama or LM Studio running