LLM EXPLORER 57,252 MODELS INDEXED

Cosmos Reason2 32B by nvidia

By nvidia · 2662 downloads
HF Repo ↗ Copy URL

Cosmos Reason2 32B is an open-source language model by nvidia. Features: 32b LLM, VRAM: 66.4GB, License: other, Instruction-Based.

  Arxiv:2312.14115   Arxiv:2406.10721   Arxiv:2503.06800   Arxiv:2603.18178 Base model:finetune:qwen/qwen3... Base model:qwen/qwen3-vl-32b-i...   Conversational   Cosmos   Image-text-to-text   Instruct   Nvidia   Qwen3 vl   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/nvidia/Cosmos-Reason2-32B 

Cosmos Reason2 32B Parameters and Internals

LLM NameCosmos Reason2 32B
Repository 🤗https://huggingface.co/nvidia/Cosmos-Reason2-32B 
Base Model(s)  Qwen3 VL 32B Instruct   Qwen/Qwen3-VL-32B-Instruct
Model Size32b
Required VRAM66.4 GB
Updated2026-08-10
Maintainernvidia
Model Typeqwen3_vl
Instruction-BasedYes
Model Files  5.2 GB: 1-of-13   5.1 GB: 2-of-13   5.3 GB: 3-of-13   5.1 GB: 4-of-13   5.1 GB: 5-of-13   5.3 GB: 6-of-13   5.1 GB: 7-of-13   5.1 GB: 8-of-13   5.3 GB: 9-of-13   5.1 GB: 10-of-13   5.1 GB: 11-of-13   5.3 GB: 12-of-13   4.3 GB: 13-of-13
Model ArchitectureQwen3VLForConditionalGeneration
Licenseother
Model Max Length262144
Transformers Version4.57.0.dev0
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Errorsreplace

Best Alternatives to Cosmos Reason2 32B

Best Alternatives
Context / RAM
Downloads
Likes
Qwen3 VL 32B Instruct0K / 67 GB2881482223
Qwen3 VL 32B Instruct FP80K / 35.7 GB38061947
Qwen3 VL 32B Instruct FP80K / 35.7 GB301
Qwen3 VL 32B Instruct0K / 67 GB3305
Qwen3 VL 32B Instruct Bnb 4bit0K / 20.5 GB2988466
...truct Ultra Uncensored Heretic0K / 66.7 GB31011
Qwen3 VL 32B Instruct 8bit0K / 36.1 GB29251
Qwen3 VL 32B Instruct 4bit0K / 19.6 GB3881
... 32B Instruct Unsloth Bnb 4bit0K / 32.5 GB14452
Qwen3 VL 32B Instruct AWQ0K / 20.7 GB54954412
Note: green Score (e.g. "73.2") means that the model is better than nvidia/Cosmos-Reason2-32B.