What are the hardware requirements for ATLAS 8B Thinking?

ATLAS 8B Thinking requires approximately 16.4 GB of VRAM and supports a context window of 40K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed ATLAS 8B Thinking and how large is it?

ATLAS 8B Thinking is developed by Arc-Intelligence, a model with 8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does ATLAS 8B Thinking perform on standard benchmarks?

ATLAS 8B Thinking has the following published scores: LMSYS ELO 1483. Compare against reference models on this page or on the LLM Explorer leaderboards.

ATLAS 8B Thinking by Arc-Intelligence — VRAM 16.4GB, 40K context

Name: ATLAS 8B Thinking
Author: Arc-Intelligence

ATLAS 8B Thinking is an open-source language model by Arc-Intelligence. Features: 8b LLM, VRAM: 16.4GB, Context: 40K, License: apache-2.0, LLM Explorer Score: 0.35, ELO: 1483.

Adaptive-learning Base model:finetune:qwen/qwen3... Base model:qwen/qwen3-8b Conversational Dataset:arc-intelligence/arc-a... Deploy:azure En Endpoints compatible Model-index Pedagogy Qwen3 Region:us Rlaif Rlhf Safetensors Sharded Teacher-student Tensorflow

Model Card on HF 🤗: https://huggingface.co/Arc-Intelligence/ATLAS-8B-Thinking

ATLAS 8B Thinking Benchmarks

LMSys ELO 1483 vs 1272 (so35) 16.6%

^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)

ATLAS 8B Thinking Parameters and Internals

LLM Name	ATLAS 8B Thinking
Repository 🤗	https://huggingface.co/Arc-Intelligence/ATLAS-8B-Thinking
Base Model(s)	Qwen3 8B Qwen/Qwen3-8B
Model Size	8b
Required VRAM	16.4 GB
Updated	2026-08-09
Maintainer	Arc-Intelligence
Model Type	qwen3
Model Files	4.9 GB: 1-of-4 4.9 GB: 2-of-4 5.0 GB: 3-of-4 1.6 GB: 4-of-4
Supported Languages	en
Model Architecture	Qwen3ForCausalLM
License	apache-2.0
Context Length	40960
Model Max Length	40960
Transformers Version	4.51.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Vocabulary Size	151936
Torch Data Type	bfloat16
Errors	replace

Rank the ATLAS 8B Thinking 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 ATLAS 8B Thinking

Best Alternatives	Context / RAM	Downloads	Likes
PaCoRe 8B	320K / 1.2 GB	54	48
RLVR 8B 0926	320K / 1.2 GB	63	8
...n3 8B 320K Context 10X Massive	320K / 16.4 GB	20	0
...r Of Horror Jan V1 256K Ctx 8B	256K / 16.1 GB	33	3
... BIG Jan Horror V1 256K Ctx 8B	256K / 16.1 GB	44	0
Qwen3 8B 256K Context 8X Grand	256K / 16.4 GB	95	0
...wen3 8B 192K Context 6X Larger	192K / 16.4 GB	55	0
DeepSeek R1 0528 Qwen3 8B	128K / 16.4 GB	1762293	1082
Marco DeepResearch 8B	128K / 16.3 GB	520	27
Marco DeepResearch 8B	128K / 16.3 GB	235	34

Note: green Score (e.g. "73.2") means that the model is better than Arc-Intelligence/ATLAS-8B-Thinking.