Skip to content
Local AI Wizard
Your machine
Theme

Cosmos-Reason2-8B

NVIDIA · nvidia/Cosmos-Reason2-8B

Weights license: Publisher terms: NVIDIA Open Model License · code and components may use other terms

Reported by the publisher16 resultsNVIDIA's model card · read 2026-10-25 · not ranked · see them
Memory needed~7 GBQ4 weights + working room, before the machine's reserve
Parameters8.8 BHub safetensors total
Repository updated2026-04-30A commit is not a release date

Quality

No comparable score

We track no public, comparable benchmark that covers this model yet. No score is inferred.

Reported by the publisher

16 results · NVIDIA's model card · read 2026-10-25
Overall
73.73General
BlinkDepth
87.90General
BlinkSpatial
84.62General
CVBench
85.60General
VideoPhy2
36.80General
Overall
56.90Robotics
ERQA
44.00Robotics
CR Common
64.24Robotics
Every reported result16
CR Embodied
69.34Robotics
Where2Place
50.00Robotics
Overall
67.85Self-Driving
AV Collision
74.00Self-Driving
AV Stop
57.14Self-Driving
LingoQA
72.40Self-Driving
Overall
69.96Smart Spaces
Warehouse AI
69.96Smart Spaces

Values as printed on the card, with the publisher's own settings (card section: Model Overview › Model Quality Performance); column “Cosmos-Reason2-8B”. Not comparable across publishers and not used in any ranking. Model card

Runs on

Fast memory holds it 90

The planning estimate fits the memory the model can use quickly: graphics card VRAM, or unified memory within the macOS GPU share.

Show all 90

Within memory, outside the fast part 26

Within the machine's memory, outside its fast part: on a graphics card the layers that do not fit run on the CPU from PC RAM (the listed RAM, or 32 GB assumed); on a Mac, above the macOS default GPU share. Not a speed figure.

Show all 26

One memory tier short 3

The same machines with the next memory size up would hold it in fast memory.

Hardware profiles grouped by the Q4 planning estimate on each machine (dedicated GPUs reserve 2 GB; unified and system memory 25%, 2 to 8 GB): the fast memory holds it, it is within memory but outside the fast part (layers offloaded to PC RAM, or above the macOS GPU share), or the machine is one memory tier short. Not a measured fit or speed.

Consider it for

NVIDIA released Cosmos-Reason2-8B on 2025-12-19: a vision-language model that answers about images, with 8.77B parameters on the Hugging Face count and the Publisher terms: NVIDIA Open Model License license.

Check first

License: commercial use allowed. At Q4 the weights load about 6.6 GB before each machine's reserve; image resolution, video frames and context add memory, and the vision encoder needs support in the runtime you pick. Downloading requires accepting the publisher's terms on Hugging Face.

Measured speed

No published measurement for this checkpoint

Tokens per second are shown only when a source publishes them with hardware, build and setup. They are never estimated.

How to run

Runtime compatibility and tokens per second have not been measured for this entry. Published metadata and a memory estimate do not prove a successful local run.

Library in source: cosmos

Documented ways to run it

The model card could not be read (missing or gated)

Choose my computer

Technical details

What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.

Checkpoint / source
nvidia/Cosmos-Reason2-8B
Task
image-text-to-text
Architecture
Qwen3VLForConditionalGeneration
Library in source
cosmos
Weight license
Publisher terms: NVIDIA Open Model LicenseCosmos-Reason2 README (nvidia-cosmos/cosmos-reason2, GitHub) · read 2026-10-18source
Commercial use
Allowed NVIDIA Open Model License Agreement, last modified 2025-10-24 (copy shipped in NVIDIA's NVlabs/GRAIL repository) · read 2026-10-18license text
Released
Cosmos-Reason2 README news: '[December 19, 2025] We have released the Cosmos-Reason2 models and code' (NVIDIA, GitHub) · read 2026-10-18source
Every stored field14 fields
Revision checked
a9fae2cf89dc64db96b12860417f0eb403013bb9
Architecture
Qwen3VLForConditionalGeneration
Library
cosmos
Hub task tag
image-text-to-text
Declared base models
Qwen/Qwen3-VL-8B-Instruct
Formats
safetensors
Access
Gated (auto)
Publisher kind
Official
Hub parameter count
8,767,123,696safetensors.total
Repository created
2025-12-12creation is not the model's release
Repository updated
2026-04-30a commit is not a release date
Hub downloads (30 days) · likes
84,651 · 245this repository only · read 2026-10-25
Read from the Hub page
task tagsHub overview and file listing · read 2026-10-19
Catalog state
Hub config or weights found; no inventory benchmark link

Sources

Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.

Curated facts checked: 2026-10-18. The license and release date cite the publisher's sources; parameters, files and task come from the Hugging Face record, and memory is a labelled estimate from its parameter count. A value no source states stays empty; a repository commit is not a model release date.