What are the hardware requirements for Kimi Linear REAP 35B A3B Instruct?

Kimi Linear REAP 35B A3B Instruct requires approximately 70.3 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Kimi Linear REAP 35B A3B Instruct and how large is it?

Kimi Linear REAP 35B A3B Instruct is developed by cerebras, a model with 48b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Kimi Linear REAP 35B A3B Instruct?

Kimi Linear REAP 35B A3B Instruct is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Kimi Linear REAP 35B A3B Instruct by cerebras — VRAM 70.3GB

Name: Kimi Linear REAP 35B A3B Instruct
Author: cerebras

Kimi Linear REAP 35B A3B Instruct is an open-source language model by cerebras. Features: 48b LLM, VRAM: 70.3GB, License: mit, Instruction-Based, LLM Explorer Score: 0.22.

Arxiv:2510.13999 Base model:finetune:moonshotai... Base model:moonshotai/kimi-lin... Compression Conversational Custom code En Instruct Kimi Kimi linear Moe Pruning Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/cerebras/Kimi-Linear-REAP-35B-A3B-Instruct

Kimi Linear REAP 35B A3B Instruct Parameters and Internals

LLM Name	Kimi Linear REAP 35B A3B Instruct
Repository 🤗	https://huggingface.co/cerebras/Kimi-Linear-REAP-35B-A3B-Instruct
Base Model(s)	Kimi Linear 48B A3B Instruct moonshotai/Kimi-Linear-48B-A3B-Instruct
Model Size	48b
Required VRAM	70.3 GB
Updated	2026-07-08
Maintainer	cerebras
Model Type	kimi_linear
Instruction-Based	Yes
Model Files	5.0 GB: 1-of-15 5.0 GB: 2-of-15 5.0 GB: 3-of-15 5.0 GB: 4-of-15 5.0 GB: 5-of-15 5.0 GB: 6-of-15 5.0 GB: 7-of-15 5.0 GB: 8-of-15 5.0 GB: 9-of-15 5.0 GB: 10-of-15 5.0 GB: 11-of-15 5.0 GB: 12-of-15 5.0 GB: 13-of-15 4.5 GB: 14-of-15 0.8 GB: 15-of-15
Supported Languages	en
Model Architecture	KimiLinearForCausalLM
License	mit
Model Max Length	1048576
Transformers Version	4.57.1
Tokenizer Class	TikTokenTokenizer
Padding Token	[PAD]
Vocabulary Size	163840

Rank the Kimi Linear REAP 35B A3B Instruct 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 Kimi Linear REAP 35B A3B Instruct

Best Alternatives	Context / RAM	Downloads	Likes
Kimi Linear 48B A3B Instruct	0K / 98.3 GB	71928	566
...near 48B A3B Instruct Mlx Bf16	0K / 98 GB	108	1
...i Linear 48B A3B Instruct 4bit	0K / 27.6 GB	1151	9
...i Linear 48B A3B Instruct 6bit	0K / 39.8 GB	390	3
...i Linear 48B A3B Instruct 8bit	0K / 52.2 GB	204	1
...r 48B A3B Instruct Abliterated	0K / 98.3 GB	122	9
...i Linear 48B A3B Instruct 3bit	0K / 21.6 GB	143	2
...i Linear 48B A3B Instruct 5bit	0K / 33.7 GB	112	1

Note: green Score (e.g. "73.2") means that the model is better than cerebras/Kimi-Linear-REAP-35B-A3B-Instruct.