What are the hardware requirements for INTELLECT 3?

INTELLECT 3 requires approximately 194.3 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed INTELLECT 3 and how large is it?

INTELLECT 3 is developed by PrimeIntellect, a model with 106.9b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does INTELLECT 3 perform on standard benchmarks?

INTELLECT 3 has the following published scores: LMSYS ELO 1356. Compare against reference models on this page or on the LLM Explorer leaderboards.

INTELLECT 3 by PrimeIntellect — VRAM 194.3GB, 128K context

Name: INTELLECT 3
Author: PrimeIntellect

INTELLECT 3 is an open-source language model by PrimeIntellect. Features: 106.9b LLM, VRAM: 194.3GB, Context: 128K, License: mit, LLM Explorer Score: 0.41, ELO: 1356.

Agentic Base model:finetune:zai-org/gl... Base model:zai-org/glm-4.5-air... Conversational Custom code En Endpoints compatible Glm4 moe Mixture-of-experts Prime-intellect Prime-rl Reasoning Region:us Safetensors Sharded Tensorflow Verifiers

Model Card on HF 🤗: https://huggingface.co/PrimeIntellect/INTELLECT-3

INTELLECT 3 Benchmarks

LMSys ELO 1356 vs 1272 (so35) 6.6%

^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)

INTELLECT 3 Parameters and Internals

LLM Name	INTELLECT 3
Repository 🤗	https://huggingface.co/PrimeIntellect/INTELLECT-3
Base Model(s)	zai-org/GLM-4.5-Air-Base zai-org/GLM-4.5-Air-Base
Model Size	106.9b
Required VRAM	194.3 GB
Updated	2026-08-08
Maintainer	PrimeIntellect
Model Type	glm4_moe
Model Files	4.9 GB: 1-of-48 4.9 GB: 2-of-48 4.5 GB: 3-of-48 4.5 GB: 4-of-48 4.5 GB: 5-of-48 4.5 GB: 6-of-48 4.5 GB: 7-of-48 4.5 GB: 8-of-48 4.5 GB: 9-of-48 4.5 GB: 10-of-48 4.5 GB: 11-of-48 4.5 GB: 12-of-48 4.5 GB: 13-of-48 4.5 GB: 14-of-48 4.5 GB: 15-of-48 4.5 GB: 16-of-48 4.5 GB: 17-of-48 4.5 GB: 18-of-48 4.5 GB: 19-of-48 4.5 GB: 20-of-48 4.5 GB: 21-of-48 4.5 GB: 22-of-48 4.5 GB: 23-of-48 4.5 GB: 24-of-48 4.5 GB: 25-of-48 4.5 GB: 26-of-48 4.5 GB: 27-of-48 4.5 GB: 28-of-48 4.5 GB: 29-of-48 4.5 GB: 30-of-48 4.5 GB: 31-of-48 4.5 GB: 32-of-48 4.5 GB: 33-of-48 4.5 GB: 34-of-48 4.5 GB: 35-of-48 4.5 GB: 36-of-48 4.5 GB: 37-of-48 4.5 GB: 38-of-48 4.5 GB: 39-of-48 4.5 GB: 40-of-48 4.5 GB: 41-of-48 4.5 GB: 42-of-48 4.5 GB: 43-of-48
Supported Languages	en
Model Architecture	Glm4MoeForCausalLM
License	mit
Context Length	131072
Model Max Length	131072
Transformers Version	4.56.1
Vocabulary Size	151552

Rank the INTELLECT 3 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

Quantized Models of the INTELLECT 3

Model	Likes	Downloads	VRAM
INTELLECT 3 4bit	3	14	60 GB
INTELLECT 3 5bit	1	6	73 GB

Best Alternatives to INTELLECT 3

Best Alternatives	Context / RAM	Downloads	Likes
GLM 4.5 Air Mxfp4	128K / 56.6 GB	58	2
INTELLECT 3 4bit	128K / 60.4 GB	14	3
INTELLECT 3 5bit	128K / 73.4 GB	6	1
GLM 4.5 Air 2bit	128K / 33.4 GB	107	4
GLM 4.5 Air 8bit	128K / 114.1 GB	2726	11
GLM 4.5 Air 4bit	128K / 60.4 GB	1261	28
GLM 4.5 Air 3bit	128K / 46.6 GB	621	32
GLM 4.5 Air 6bit	128K / 86.7 GB	201	3
GLM 4.5 Air 3bit DWQ V2	128K / 46.6 GB	82	4
GLM 4.5 Air 3bit DWQ	128K / 53.3 GB	50	4