MLX Model Test Report

Generated 2026-02-27 11:03 · AFM MLX Backend · v0.9.5-4d58e3f

mlx-model-test.sh --model mlx-community/Qwen3.5-35B-A3B-4bit --prompts /Volumes/edata/dev/git/NIGHTLY/maclocal-api/Scripts/test-llm-comprehensive.txt --tests 4,7,23 --smart 1:claude

Test Runs
3
Passed
3
Failed
0
Best tok/s
112.3
Fastest
mlx-community/Qwen3.5-35B-A3B-4bit @ top-p

Performance Ranking (by tokens/sec)

Click a row to jump to its full response below.

# Model / Config Status claude Temp Load (s) Tokens Gen (s) Tokens/sec Prompt
1 mlx-community/Qwen3.5-35B-A3B-4bit @ top-p top-p
--top-p 0.9		 OK 5 0.8 2.0 943 8.4
112.3
 Explain why the sky is blue in exactly 3 bullet points.
2 mlx-community/Qwen3.5-35B-A3B-4bit @ combined-samplers combined-samplers
--top-k 50 --min-p 0.03 --top-p 0.95		 OK 5 0.8 1.0 848 7.75
109.4
 Explain why the sky is blue in exactly 3 bullet points.
3 mlx-community/Qwen3.5-35B-A3B-4bit @ agent-cached-turn1 agent-cached-turn1
--enable-prefix-caching		 OK 4 0.0 1.0 67 1.15
58.0
 Read the file Sources/MacLocalAPI/main.swift and explain wha...

AI Analysis (--smart)

▶ Quality, Anomalies & Recommendations · avg score: 4.2/5

Full Responses

▶ mlx-community/Qwen3.5-35B-A3B-4bit @ top-p 943 tokens · 112.3 tok/s

▶ mlx-community/Qwen3.5-35B-A3B-4bit @ combined-samplers 848 tokens · 109.4 tok/s

▶ mlx-community/Qwen3.5-35B-A3B-4bit @ agent-cached-turn1 67 tokens · 58.0 tok/s