qwen3:4b

OLLAMA GGUF

qwen3 · 4.0B · Q4_K_M

Not Rec.

Jun 26, 2026 · Cortex-A78AE

deepseek-v2:16b

OLLAMA GGUF

deepseek2 · 15.7B · Q4_0

Good

Mar 2, 2026 · Apple M4

Global Score

29 vs 69

Hardware Fit

86 vs 100

Quality Score

4 vs 55

Hardware

qwen3:4b deepseek-v2:16b

MachineNVIDIA NVIDIA Jetson Orin Nano Engineering Reference Developer Kit SuperMacBook Air

CPUCortex-A78AEApple M4

Cores610

RAM7 GB32 GB

GPUIntegrated / UnknownApple M4

OSUbuntu 22.04.5 LTSmacOS 26.3

Archarm64arm64

Power Modebalancedbalanced

Performance

qwen3:4b deepseek-v2:16b

Tokens/sec17.956.6

First chunk1010 msN/A

TTFT6.4 s239 ms

Load time9.5 s1.2 s

Memory usage3.0 GB17.7 GB

Memory %40%55%

HW Fit Score Breakdown

qwen3:4b

Speed

50/50

TTFT

8/20

Memory

28/30

deepseek-v2:16b

Speed

40/50

TTFT

30/20

Memory

30/30

Quality

qwen3:4b

Reasoning

4/20

Coding

0/20

Instruction

0/20

Structured

0/15

Math

0/15

Multilingual

0/10

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

deepseek-v2:16b

Reasoning

7/20

Coding

9/20

Instruction

13/20

Structured

10/15

Math

8/15

Multilingual

8/10

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

View qwen3:4b details | View deepseek-v2:16b details

Run yours and compare

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest