LLM EXPLORER 59,358 MODELS INDEXED

Llama 4 Maverick 17B 16E Instruct 4bit by mlx-community

By mlx-community · 419 downloads
HF Repo ↗ Copy URL

Llama 4 Maverick 17B 16E Instruct 4bit is an open-source language model by mlx-community. Features: 17b LLM, VRAM: 146.3GB, License: other, Quantized, Instruction-Based, LLM Explorer Score: 0.31, ELO: 1327.

  4-bit   4bit   Ar Base model:meta-llama/llama-4-... Base model:quantized:meta-llam...   Conversational   De   En   Es   Facebook   Fr   Hi   Id   Instruct   It   Llama   Llama-4   Llama4   Meta   Mlx   Pt   Pytorch   Quantized   Region:us   Safetensors   Sharded   Tensorflow   Th   Tl   Vi
Model Card on HF 🤗: https://huggingface.co/mlx-community/Llama-4-Maverick-17B-16E-Instruct-4bit 

Llama 4 Maverick 17B 16E Instruct 4bit Benchmarks

LMSys ELO 1327 vs 1272 (so35) 4.3%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

Llama 4 Maverick 17B 16E Instruct 4bit Parameters and Internals

LLM NameLlama 4 Maverick 17B 16E Instruct 4bit
Repository 🤗https://huggingface.co/mlx-community/Llama-4-Maverick-17B-16E-Instruct-4bit 
Base Model(s)  meta-llama/Llama-4-Maverick-17B-128E-Instruct   meta-llama/Llama-4-Maverick-17B-128E-Instruct
Model Size17b
Required VRAM146.3 GB
Updated2026-07-25
Maintainermlx-community
Model Typellama4
Instruction-BasedYes
Model Files  3.8 GB: 1-of-72   3.0 GB: 2-of-72   3.3 GB: 3-of-72   3.0 GB: 4-of-72   3.0 GB: 5-of-72   3.3 GB: 6-of-72   3.0 GB: 7-of-72   3.0 GB: 8-of-72   3.3 GB: 9-of-72   3.0 GB: 10-of-72   3.0 GB: 11-of-72   3.3 GB: 12-of-72   3.0 GB: 13-of-72   3.0 GB: 14-of-72   3.3 GB: 15-of-72   3.0 GB: 16-of-72   3.0 GB: 17-of-72   3.3 GB: 18-of-72   3.0 GB: 19-of-72   3.0 GB: 20-of-72   3.3 GB: 21-of-72   3.0 GB: 22-of-72   3.0 GB: 23-of-72   3.3 GB: 24-of-72   3.0 GB: 25-of-72   3.0 GB: 26-of-72   3.3 GB: 27-of-72   3.0 GB: 28-of-72   3.0 GB: 29-of-72   3.3 GB: 30-of-72   3.0 GB: 31-of-72   3.0 GB: 32-of-72   3.3 GB: 33-of-72   3.0 GB: 34-of-72   3.0 GB: 35-of-72   3.3 GB: 36-of-72   3.0 GB: 37-of-72   3.0 GB: 38-of-72   3.3 GB: 39-of-72   3.0 GB: 40-of-72   3.0 GB: 41-of-72   3.3 GB: 42-of-72   3.0 GB: 43-of-72   3.0 GB: 44-of-72   3.3 GB: 45-of-72   3.0 GB: 46-of-72   3.0 GB: 47-of-72
Supported Languagesar de en es fr hi id it pt th tl vi
Quantization Type4bit
Model ArchitectureLlama4ForConditionalGeneration
Licenseother
Transformers Version4.51.0.dev0
Torch Data Typebfloat16

Best Alternatives to Llama 4 Maverick 17B 16E Instruct 4bit

Best Alternatives
Context / RAM
Downloads
Likes
... 4 Scout 17B 16E Instruct 6bit0K / 88.8 GB3335
...averick 17B 128E Instruct 6bit0K / 201.1 GB1251
... 4 Scout 17B 16E Instruct 4bit0K / 61 GB120312
... 4 Scout 17B 16E Instruct 8bit0K / 114.7 GB5884
... 16E Instruct Unsloth Bnb 4bit0K / 60.6 GB20580
...averick 17B 128E Instruct 4bit0K / 140.8 GB2042
...truct Unsloth Dynamic Bnb 4bit0K / 61.7 GB1367
...Maverick 17B 16E Instruct 6bit0K / 213.8 GB1312
... 16E Instruct Unsloth Bnb 8bit0K / 111.5 GB229
Llama 4 Scout 17B 16E Instruct0K / 188.7 GB358657
Note: green Score (e.g. "73.2") means that the model is better than mlx-community/Llama-4-Maverick-17B-16E-Instruct-4bit.