qwen3.5:35b-a3b-q4_K_M

qwen35moe · 36.0B · Q4_K_M

THINKING MODEL

Dell Inc. Precision Tower 7910 (Intel Xeon® E5-2637 v3)

96 GB · Microsoft Windows 10 Pro 10.0.19045

Tested on July 30, 2026 · Submitted by Csaba

Top 87% Compare

Global Score

49 /100

Not Rec.

Hardware Fit

54/100

Quality

47/100

Get this model

🦙

Ollama

ollama pull qwen3.5:35b-a3b-q4_K_M

View on Ollama Library

ollama.com/library/qwen3.5

Get it in LM Studio

Search and download models directly from the app

🤗

Find on HuggingFace

GGUF versions & quantizations

Hardware

Machine: Dell Inc. Precision Tower 7910
CPU: Intel Xeon® E5-2637 v3
Cores: 16 total (16 perf)
Frequency: 3.5 GHz
RAM: 96 GB DDR4
GPU: NVIDIA GeForce RTX 5060 Ti
OS: Microsoft Windows 10 Pro 10.0.19045
Arch: x64
Power Mode: balanced

Performance

Tokens/sec Tokens generated per second — higher is better: 21.6
Standard deviation Variation of speed across benchmark runs — lower means more consistent: ±3.3
First chunk latency Delay before the first streamed chunk arrives from runtime: 1.5 s
Time to first token How fast the model starts responding — lower is better: 30.0 s
Load time Time to load the model into memory before inference: 53.8 s
Memory usage RAM consumed during inference — percentage of total system memory: 22.7 GB (24%)
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: ~780

Score breakdown

Speed

24/50

Time to first token

0/20

Memory

30/30

Quality

Reasoning

14/20

Coding

1/20

Instruction following

6/20

Structured output

5/15

Math

11/15

Multilingual

10/10

Category levels

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

Metadata

Spec version: 0.2.1
Runtime: Ollama 0.32.1
Model format: GGUF
Hardware profile: HIGH-END
Result hash: dec36352b5527ca11f5870f86a18ec3ff67b76003c47afb9dde3c38640f61842

Interpretation

Hardware fit: 54/100. Overall suitability: NOT RECOMMENDED (Global 49/100). Category profile: Reasoning: Adequate, Coding: Poor, Instruction Following: Weak, Structured Output: Weak, Math: Adequate, Multilingual: Strong.

Disqualifiers

Time to first token too high: 30000ms (maximum: 13009ms for HIGH-END profile)

Bench Environment

CPU load: avg 31% (peak 40%)

Run yours now

$ npm install -g metrillm@latest

$ metrillm

Requires Node 20+ and Ollama or LM Studio running

Or run without installing: npx metrillm@latest