What are the hardware requirements for Hy3 Preview 8bit?

Hy3 Preview 8bit requires approximately 235 GB of VRAM and supports a context window of 256K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Hy3 Preview 8bit and how large is it?

Hy3 Preview 8bit is developed by mlx-community, a model with 295b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Hy3 Preview 8bit perform on standard benchmarks?

Hy3 Preview 8bit has the following published scores: LMSYS ELO 1412. Compare against reference models on this page or on the LLM Explorer leaderboards.

Hy3 Preview 8bit by mlx-community — VRAM 235GB, 256K context

Name: Hy3 Preview 8bit
Author: mlx-community

Hy3 Preview 8bit is an open-source language model by mlx-community. Features: 295b LLM, VRAM: 235GB, Context: 256K, License: other, Quantized, LLM Explorer Score: 0.38, ELO: 1412.

8-bit 8bit Base model:quantized:tencent/h... Base model:tencent/hy3-preview Conversational Hy v3 Mlx Quantized Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/Hy3-preview-8bit

Hy3 Preview 8bit Benchmarks

LMSys ELO 1412 vs 1272 (so35) 11%

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

What is the LLM Explorer Rank (Score)

Hy3 Preview 8bit Parameters and Internals

LLM Name	Hy3 Preview 8bit
Repository 🤗	https://huggingface.co/mlx-community/Hy3-preview-8bit
Base Model(s)	Hy3 Preview tencent/Hy3-preview
Model Size	295b
Required VRAM	235 GB
Updated	2026-08-10
Maintainer	mlx-community
Model Type	hy_v3
Model Files	4.8 GB: 1-of-60 5.2 GB: 2-of-60 5.2 GB: 3-of-60 5.3 GB: 4-of-60 5.2 GB: 5-of-60 5.2 GB: 6-of-60 5.3 GB: 7-of-60 5.2 GB: 8-of-60 5.2 GB: 9-of-60 5.3 GB: 10-of-60 5.2 GB: 11-of-60 5.2 GB: 12-of-60 5.3 GB: 13-of-60 5.2 GB: 14-of-60 5.2 GB: 15-of-60 5.3 GB: 16-of-60 5.2 GB: 17-of-60 5.2 GB: 18-of-60 5.3 GB: 19-of-60 5.2 GB: 20-of-60 5.2 GB: 21-of-60 5.3 GB: 22-of-60 5.2 GB: 23-of-60 5.2 GB: 24-of-60 5.3 GB: 25-of-60 5.2 GB: 26-of-60 5.2 GB: 27-of-60 5.3 GB: 28-of-60 5.2 GB: 29-of-60 5.2 GB: 30-of-60 5.3 GB: 31-of-60 5.2 GB: 32-of-60 5.2 GB: 33-of-60 5.3 GB: 34-of-60 5.2 GB: 35-of-60 5.2 GB: 36-of-60 5.3 GB: 37-of-60 5.2 GB: 38-of-60 5.2 GB: 39-of-60 5.3 GB: 40-of-60 5.2 GB: 41-of-60 5.2 GB: 42-of-60 5.3 GB: 43-of-60 5.2 GB: 44-of-60 5.2 GB: 45-of-60
Quantization Type	8bit
Model Architecture	HYV3ForCausalLM
License	other
Context Length	262144
Model Max Length	262144
Transformers Version	5.6.0
Vocabulary Size	120832

Rank the Hy3 Preview 8bit capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Hy3 Preview 8bit

Best Alternatives	Context / RAM	Downloads	Likes
Hy3 Preview 4bit	256K / 166.5 GB	1563	3
Hy3 Preview 6bit	256K / 227.6 GB	93	1
Hy3 Heretic	256K / 323.1 GB	324	6
Hy3 GPTQ RTN 4bit Tp8	256K / 161.3 GB	46	2

Note: green Score (e.g. "73.2") means that the model is better than mlx-community/Hy3-preview-8bit.