MLX Model Test Report

Generated 2026-02-25 21:38 · AFM MLX Backend · v0.9.5-0cfba17

mlx-model-test.sh --prompts Scripts/test-vlm-media.txt

Test Runs
7
Passed
5
Failed
2
Best tok/s
40.8
Fastest
mlx-community/Qwen3.5-35B-A3B-4bit @ image-greedy

Performance Ranking (by tokens/sec)

Click a row to jump to its full response below.

# Model / Config Status Temp Load (s) Tokens Gen (s) Tokens/sec Prompt
1 mlx-community/Qwen3.5-35B-A3B-4bit @ image-greedy image-greedy
--vlm
OK 0.0 1.0 4096 100.5
40.8
What is in this image? Reply in one sentence.
2 mlx-community/Qwen3.5-35B-A3B-4bit @ image-system-prompt image-system-prompt
--vlm
OK 0.7 1.0 1042 28.52
36.5
Describe the animal in this image. What breed might it be?
3 mlx-community/Qwen3.5-35B-A3B-4bit @ image-describe image-describe
--vlm
OK 0.7 1.0 918 25.55
35.9
Describe this image in detail.
4 mlx-community/Qwen3.5-35B-A3B-4bit @ image-question image-question
--vlm
OK 0.7 1.0 220 9.25
23.8
What animal is in this image? Answer in one word.
5 mlx-community/Qwen3.5-35B-A3B-4bit @ image-stop image-stop
--vlm
OK 0.7 1.0 16 4.33
3.7
Describe this image in detail

Failed Runs

ModelErrorConfigLoad (s)
mlx-community/Qwen3.5-35B-A3B-4bit @ image-guided-json image-guided-json Error code: 400 - {'error': {'message': 'The operation couldn’t be completed. (Jinja.TemplateException error 1.)', 'type': 'mlx_error'}} t=0.7 --vlm 1.0
mlx-community/Qwen3.5-35B-A3B-4bit @ video-describe video-describe Error code: 400 - {'error': {'type': 'mlx_error', 'message': 'Unable to load image from URL: video/mp4;base64,AAAAIGZ0eXBpc29tAAACAGlzb21pc28yYXZjMW1wNDEAAAAIZnJlZQAPNwdtZGF0AAACrwYF//+r3EXpvebZSLeWLN t=0.7 --vlm 1.0

Full Responses

mlx-community/Qwen3.5-35B-A3B-4bit @ image-greedy 4096 tokens · 40.8 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-system-prompt 1042 tokens · 36.5 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-describe 918 tokens · 35.9 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-question 220 tokens · 23.8 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-stop 16 tokens · 3.7 tok/s