MLX Model Test Report

Generated 2026-02-26 11:09 · AFM MLX Backend · v0.9.5

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

Test Runs
10
Passed
10
Failed
0
Best tok/s
89.3
Fastest
mlx-community/Qwen3.5-35B-A3B-4bit @ llm-creative

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 @ llm-creative llm-creative OK 0.7 2.0 512 5.73
89.3
Write a haiku about the ocean.
2 mlx-community/Qwen3.5-35B-A3B-4bit @ image-ocr image-ocr
--vlm
OK 0.7 2.0 2745 30.86
89.0
Read all the text in this image. Output it exactly as writte...
3 mlx-community/Qwen3.5-35B-A3B-4bit @ llm-factual llm-factual OK 0.7 1.0 210 2.45
85.6
What is the capital of Japan? Answer in one sentence.
4 mlx-community/Qwen3.5-35B-A3B-4bit @ image-system-prompt image-system-prompt
--vlm
OK 0.7 1.0 1077 13.36
80.6
Describe the animal in this image. What breed might it be?
5 mlx-community/Qwen3.5-35B-A3B-4bit @ image-stop image-stop
--vlm
OK 0.7 2.0 701 9.39
74.7
Describe this image in detail
6 mlx-community/Qwen3.5-35B-A3B-4bit @ image-question image-question
--vlm
OK 0.7 1.0 470 6.48
72.5
What animal is in this image? Answer in one word.
7 mlx-community/Qwen3.5-35B-A3B-4bit @ image-greedy image-greedy
--vlm
OK 0.0 2.0 472 6.57
71.8
What is in this image? Reply in one sentence.
8 mlx-community/Qwen3.5-35B-A3B-4bit @ image-describe image-describe
--vlm
OK 0.7 4.0 1084 17.99
60.3
Describe this image in detail.
9 mlx-community/Qwen3.5-35B-A3B-4bit @ video-describe video-describe
--vlm
OK 0.7 2.0 848 15.07
56.3
Describe what happens in this video.
10 mlx-community/Qwen3.5-35B-A3B-4bit @ image-guided-json image-guided-json
--vlm
OK 0.7 1.0 1305 24.9
52.4
Analyze this image and return a JSON object with the followi...

Full Responses

mlx-community/Qwen3.5-35B-A3B-4bit @ llm-creative 512 tokens · 89.3 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-ocr 2745 tokens · 89.0 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ llm-factual 210 tokens · 85.6 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-system-prompt 1077 tokens · 80.6 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-stop 701 tokens · 74.7 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-question 470 tokens · 72.5 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-greedy 472 tokens · 71.8 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-describe 1084 tokens · 60.3 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ video-describe 848 tokens · 56.3 tok/s

mlx-community/Qwen3.5-35B-A3B-4bit @ image-guided-json 1305 tokens · 52.4 tok/s