What are the hardware requirements for Step 3.5 Flash 6bit?

Step 3.5 Flash 6bit requires approximately 159.9 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 Step 3.5 Flash 6bit and how large is it?

Step 3.5 Flash 6bit is developed by mlx-community, a model with 197b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Step 3.5 Flash 6bit perform on standard benchmarks?

Step 3.5 Flash 6bit has the following published scores: LMSYS ELO 1395. Compare against reference models on this page or on the LLM Explorer leaderboards.

Step 3.5 Flash 6bit by mlx-community — VRAM 159.9GB, 256K context

Name: Step 3.5 Flash 6bit
Author: mlx-community

Step 3.5 Flash 6bit is an open-source language model by mlx-community. Features: 197b LLM, VRAM: 159.9GB, Context: 256K, License: apache-2.0, Quantized, LLM Explorer Score: 0.37, ELO: 1395.

6-bit 6bit Base model:quantized:stepfun-a... Base model:stepfun-ai/step-3.5... Conversational Custom code Mlx Quantized Region:us Safetensors Sharded Step3p5 Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlx-community/Step-3.5-Flash-6bit

Step 3.5 Flash 6bit Benchmarks

LMSys ELO 1395 vs 1272 (so35) 9.7%

^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)

Step 3.5 Flash 6bit Parameters and Internals

LLM Name	Step 3.5 Flash 6bit
Repository 🤗	https://huggingface.co/mlx-community/Step-3.5-Flash-6bit
Base Model(s)	stepfun-ai/Step-3.5-Flash stepfun-ai/Step-3.5-Flash
Model Size	197b
Required VRAM	159.9 GB
Updated	2026-08-11
Maintainer	mlx-community
Model Type	step3p5
Model Files	4.8 GB: 1-of-32 5.0 GB: 2-of-32 5.0 GB: 3-of-32 5.1 GB: 4-of-32 5.0 GB: 5-of-32 5.0 GB: 6-of-32 5.1 GB: 7-of-32 5.0 GB: 8-of-32 5.0 GB: 9-of-32 5.1 GB: 10-of-32 5.0 GB: 11-of-32 5.0 GB: 12-of-32 5.1 GB: 13-of-32 5.0 GB: 14-of-32 5.0 GB: 15-of-32 5.1 GB: 16-of-32 5.0 GB: 17-of-32 5.0 GB: 18-of-32 5.1 GB: 19-of-32 5.0 GB: 20-of-32 5.0 GB: 21-of-32 5.1 GB: 22-of-32 5.0 GB: 23-of-32 5.0 GB: 24-of-32 5.1 GB: 25-of-32 5.0 GB: 26-of-32 5.0 GB: 27-of-32 5.1 GB: 28-of-32 5.0 GB: 29-of-32 5.0 GB: 30-of-32 5.1 GB: 31-of-32 4.1 GB: 32-of-32
Quantization Type	6bit
Model Architecture	Step3p5ForCausalLM
License	apache-2.0
Context Length	262144
Model Max Length	262144
Tokenizer Class	TokenizersBackend
Padding Token	<｜end▁of▁sentence｜>
Vocabulary Size	128896
Torch Data Type	bfloat16

Rank the Step 3.5 Flash 6bit 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 Step 3.5 Flash 6bit

Best Alternatives	Context / RAM	Downloads	Likes
Step 3.5 Flash 8bit	256K / 210.7 GB	263	2
Step 3.5 Flash 4bit	256K / 110.3 GB	1164	10
Step 3.5 Flash Bf16	256K / 138.5 GB	48	1

Note: green Score (e.g. "73.2") means that the model is better than mlx-community/Step-3.5-Flash-6bit.