MLX Model Test Report

Generated 2026-02-18 22:25 · AFM MLX Backend · Apple M3 Ultra (512GB) · 3000 max tokens · temp 0.7

Models Tested
31
Passed
27
Failed
4
Best tok/s
355.5
Fastest Model
mlx-community/Llama-3.2-1B-Instruct-4bit

Performance Ranking (by tokens/sec)

Click a row to jump to its full response below.

# Model Status Load (s) Tokens Gen Time (s) Tokens/sec Response Preview
1 mlx-community/Llama-3.2-1B-Instruct-4bit OK 2 849 2.39
355.5
 Limits through Multivariable Calculus: An Introduction to Mathematical Analysis **Introduction** Calculus is a branch
2 mlx-community/Qwen2.5-0.5B-Instruct-4bit OK 1 502 1.56
321.6
 Calculus is a branch of mathematics that deals with the study of functions and their derivatives. The concept of limits
3 mlx-community/Qwen3-0.6B-4bit OK 2 2757 9.87
279.2
 <think> Okay, let me try to explain calculus concepts using limits and multivariable calculus. First, I need to re
4 mlx-community/LFM2-VL-3B-4bit OK 1 669 2.67
250.3
 Calculus is a branch of mathematics that deals with continuous change. It is divided into two main parts: differential c
5 mlx-community/exaone-4.0-1.2b-4bit OK 1 677 3.89
173.9
 Certainly! Below is a rigorous explanation of key concepts in calculus, from limits to multivariable calculus, using pre
6 mlx-community/LFM2-2.6B-4bit OK 1 2099 12.62
166.3
 Certainly! Let us explore the foundational concepts of calculus—starting from **limits**, and gradually build up to **mu
7 mlx-community/lille-130m-instruct-8bit OK 1 586 3.67
159.6
 Calculus is a subfield of calculus that deals with the study of limits, which is a fundamental concept in calculus. It i
8 mlx-community/Ling-mini-2.0-4bit OK 3 2121 15.07
140.8
 Certainly! Let's walk through **calculus concepts**, starting from **limits** and progressing through **single-vari
9 mlx-community/SmolLM3-3B-4bit OK 2 1278 12.4
103.1
 <think> </think> Let's explore calculus concepts from limits through multivariable calculus, using rig
10 mlx-community/granite-4.0-h-tiny-4bit OK 2 1012 11.09
91.3
 Certainly! Here's an explanation of calculus concepts from limits through multivariable calculus using rigorous mat
11 hub/models--mlx-community--Qwen3-VL-4B-Instruct-4bit OK 2 1982 23.63
83.9
 Certainly. Below is a rigorous, mathematically precise exposition of calculus concepts, progressing from **limits** thro
12 mlx-community/Qwen3-VL-4B-Instruct-4bit OK 2 1976 23.65
83.5
 Certainly. Below is a rigorous, mathematically precise exposition of calculus concepts, progressing from limits through
13 mlx-community/gemma-3-4b-it-8bit OK 3 1313 16.21
81.0
 Okay, let's embark on a journey through calculus, starting with the foundational limits and progressing to multivar
14 mlx-community/gpt-oss-20b-MXFP4-Q4 OK 4 1882 24.38
77.2
 <|channel|>analysis<|message|>The user requests: "Explain calculus concepts from limits through multiva
15 mlx-community/Qwen3-30B-A3B-4bit OK 6 2317 31.86
72.7
 <think> Okay, the user wants an explanation of calculus concepts from limits through multivariable calculus with r
16 mlx-community/gpt-oss-20b-MXFP4-Q8 OK 3 1965 27.36
71.8
 <|channel|>analysis<|message|>We need to explain calculus concepts from limits through multivariable calculu
17 mlx-community/gemma-3n-E2B-it-lm-4bit OK 3 2095 30.45
68.8
 Okay, let's embark on a journey through the fascinating world of calculus! I'll break down the core concepts,
18 mlx-community/Qwen3-VL-4B-Instruct-8bit OK 2 2104 32.49
64.8
 Certainly. Below is a rigorous, mathematically precise exposition of calculus concepts, progressing from limits through
19 mlx-community/GLM-4.7-Flash-4bit OK 7 2363 46.83
50.5
 The user wants a comprehensive explanation of calculus concepts, ranging from limits to multivariable calculus, using ri
20 mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit OK 6 2367 50.0
47.3
 The user wants an explanation of calculus concepts from limits through multivariable calculus with rigorous mathematical
21 mlx-community/Qwen3-Coder-Next-4bit OK 14 1860 47.89
38.8
 # A Rigorous Tour Through Calculus: From Limits to Multivariable Calculus ## 1. Limits ### 1.1 ε-δ Definition of a Lim
22 mlx-community/JoyAI-LLM-Flash-4bit-DWQ OK 9 1244 33.25
37.4
 I'll provide a rigorous explanation of calculus concepts from limits through multivariable calculus, using proper m
23 mlx-community/MiniMax-M2.5-6bit OK 59 1780 61.47
29.0
 Hmm, this is a comprehensive request covering the entire calculus sequence. The user wants rigorous notation, so I need
24 mlx-community/MiniMax-M2.5-5bit OK 48 2106 73.43
28.7
 We are going to explain calculus concepts from limits through multivariable calculus with rigorous mathematical notation
25 mlx-community/Apertus-8B-Instruct-2509-4bit OK 2 17 0.65
26.2
 keteMn}s dlůmõ okallweitev1.stl Vàvletů (vazl machineύþ�فيлаούộظàl
26 mlx-community/Qwen3.5-397B-A17B-4bit OK 68 2282 92.55
24.7
 Here's a thinking process that leads to the suggested explanation of calculus concepts: 1. **Analyze the Request:
27 mlx-community/GLM-5-4bit OK 126 2322 227.4
10.2
 The user wants a comprehensive explanation of calculus concepts, starting from limits and progressing through multivaria

Failed Models

ModelErrorLoad Time (s)
lmstudio-community/gemma-3n-E4B-it-MLX-4bit lmstudio-community/gemma-3n-E4B-it-MLX-4bit: Unsupported model type: gemma3n 300
mlx-community/Mixtral-8x7B-Instruct-v0.1-4bit mlx-community/Mixtral-8x7B-Instruct-v0.1-4bit: Unsupported model type: mixtral 300
sentence-transformers/all-MiniLM-L6-v2 sentence-transformers/all-MiniLM-L6-v2: Unsupported model type: bert 300
mlx-community/Kimi-K2.5-3bit 144

Full Model Responses

▶ mlx-community/Llama-3.2-1B-Instruct-4bit 849 tokens · 355.5 tok/s

▶ mlx-community/Qwen2.5-0.5B-Instruct-4bit 502 tokens · 321.6 tok/s

▶ mlx-community/Qwen3-0.6B-4bit 2757 tokens · 279.2 tok/s

▶ mlx-community/LFM2-VL-3B-4bit 669 tokens · 250.3 tok/s

▶ mlx-community/exaone-4.0-1.2b-4bit 677 tokens · 173.9 tok/s

▶ mlx-community/LFM2-2.6B-4bit 2099 tokens · 166.3 tok/s

▶ mlx-community/lille-130m-instruct-8bit 586 tokens · 159.6 tok/s

▶ mlx-community/Ling-mini-2.0-4bit 2121 tokens · 140.8 tok/s

▶ mlx-community/SmolLM3-3B-4bit 1278 tokens · 103.1 tok/s

▶ mlx-community/granite-4.0-h-tiny-4bit 1012 tokens · 91.3 tok/s

▶ hub/models--mlx-community--Qwen3-VL-4B-Instruct-4bit 1982 tokens · 83.9 tok/s

▶ mlx-community/Qwen3-VL-4B-Instruct-4bit 1976 tokens · 83.5 tok/s

▶ mlx-community/gemma-3-4b-it-8bit 1313 tokens · 81.0 tok/s

▶ mlx-community/gpt-oss-20b-MXFP4-Q4 1882 tokens · 77.2 tok/s

▶ mlx-community/Qwen3-30B-A3B-4bit 2317 tokens · 72.7 tok/s

▶ mlx-community/gpt-oss-20b-MXFP4-Q8 1965 tokens · 71.8 tok/s

▶ mlx-community/gemma-3n-E2B-it-lm-4bit 2095 tokens · 68.8 tok/s

▶ mlx-community/Qwen3-VL-4B-Instruct-8bit 2104 tokens · 64.8 tok/s

▶ mlx-community/GLM-4.7-Flash-4bit 2363 tokens · 50.5 tok/s

▶ mlx-community/NVIDIA-Nemotron-3-Nano-30B-A3B-4bit 2367 tokens · 47.3 tok/s

▶ mlx-community/Qwen3-Coder-Next-4bit 1860 tokens · 38.8 tok/s

▶ mlx-community/JoyAI-LLM-Flash-4bit-DWQ 1244 tokens · 37.4 tok/s

▶ mlx-community/MiniMax-M2.5-6bit 1780 tokens · 29.0 tok/s

▶ mlx-community/MiniMax-M2.5-5bit 2106 tokens · 28.7 tok/s

▶ mlx-community/Apertus-8B-Instruct-2509-4bit 17 tokens · 26.2 tok/s

▶ mlx-community/Qwen3.5-397B-A17B-4bit 2282 tokens · 24.7 tok/s

▶ mlx-community/GLM-5-4bit 2322 tokens · 10.2 tok/s