Back to leaderboard

microsoft/phi-4-mini-reasoning

LM-STUDIO MLX

phi3 · 3.8B · 4bit

Good

Mar 5, 2026 · Apple M4 Pro

qwen3-235b-a22b-t48-c8k:latest

OLLAMA GGUF

qwen3moe · 235.1B · Q4_K_M

Not Rec.

Sep 7, 2026 · Intel Xeon® E5-4650 v2

Global Score
64 vs 65
Hardware Fit
100 vs 29
Quality Score
48 vs 81

Hardware

microsoft/phi-4-mi… qwen3-235b-a22b-t4…
MachineMac miniDell Inc. PowerEdge R820
CPUApple M4 ProIntel Xeon® E5-4650 v2
Cores1480
RAM64 GB256 GB
GPUApple M4 ProMicrosoft Remote Display Adapter, NVIDIA Tesla P4, NVIDIA Tesla P4, Microsoft Basic Display Adapter
OSmacOS 15.7.4Microsoft Windows Server 2022 Datacenter 10.0.20348
Archarm64x64
Power Modebalancedbalanced

Performance

microsoft/phi-4-mi… qwen3-235b-a22b-t4…
Tokens/sec90.32.4
First chunk198 ms5833 ms
TTFT198 ms5.8 s
Load timeN/A0.0 s
Memory usage2.0 GB135.1 GB
Memory %3%53%

HW Fit Score Breakdown

microsoft/phi-4-mini-re…

Speed
50/50
TTFT
20/20
Memory
30/30

qwen3-235b-a22b-t48-c8k…

Speed
2/50
TTFT
3/20
Memory
24/30

Quality

microsoft/phi-4-mini-re…

Reasoning
7/20
Coding
7/20
Instruction
7/20
Structured
7/15
Math
13/15
Multilingual
7/10
Reasoning: Weak Coding: Weak Instruction Following: Weak Structured Output: Weak Math: Strong Multilingual: Adequate

qwen3-235b-a22b-t48-c8k…

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

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