NVLM-H 72B
No estimate
No hardware requirements for this model
The weights for this model have not been published, so it cannot be downloaded or run on your own hardware at any size. It is reachable only through its provider, and no graphics card changes that.
On record
Full specification
Everything on record for this model. Most of it describes how it was trained rather than how it runs — useful context for judging how much work went into it, and how it compares with models built at a different scale.
Origin
Who built this model, where, and when it was published.
- Organisation
- NVIDIA
- Organisation type
- Industry
- Country
- United States of America
- Published
- 22 October 2024
- Authors
- Wenliang Dai, Nayeon Lee, Boxin Wang, Zhuolin Yang, Zihan Liu, Jon Barker, Tuomas Rintamaki, Mohammad Shoeybi, Bryan Catanzaro, Wei Ping
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Vision, Language
- Task
- Language modeling/generation, Vision-language generation, Question answering, Code generation, Translation, Quantitative reasoning
- Base model
- Qwen2-72B,InternViT-6B
- Numerical format
- BF16
Size
How large the model is and how much data it was trained on. Parameters are the figure that decides whether it fits on a given graphics card.
- Parameters
- 72B
- Training data
- 125,829,120,000 tokens
- Epochs
- 1
72B
Pre-training: Global batch size 2,048 Sequence length in the LLM decoder 512 Downsampling of visual tokens 1024->256 # of visual token per tile 256 # of tiles 6+1 # of training steps 20K 2048 * (512+256*7) * 20000 = 94,371,840,000 SFT: Global batch size 256 Sequence length in the LLM decoder 1,280 # of visual token per tile 256 # of tiles 6+1 # of training steps 40K 256*(1280+256*7)*40000 = 31,457,280,000 94,371,840,000 + 31,457,280,000 = 125,829,120,000
Training compute
The arithmetic performed to train the model, measured in floating-point operations. It is a measure of what the training run cost, not of how fast the finished model answers you.
- Training compute
- 3 × 10²⁴ FLOP
- How it was established
- Operation counting
- Fine-tuning compute
- 5.4 × 10²² FLOP
Additional compute in this paper is negligible relative to the compute used to train the language model backbone (Qwen2-72B at 3.02e24 FLOP)
6ND = 6*125,829,120,000*72000000000.00 = 5.436e22
The training run
What it physically took to train: which chips, how many, for how long, and what that drew from the wall.
- Training hardware
- NVIDIA H100 SXM5 80GB
- Chips used
- 128
- Power draw
- 176.4 kW
Availability
Whether you can obtain the model and run it on your own hardware, which is what decides if any of the graphics-card figures on this page apply.
- Weights
- Closed — provider access only
- Model access
- Unreleased
- Training code
- Unreleased
How it is classified
Labels the source dataset applies when tracking notable models, and how confident it is in the entry.
- Likely above 10²³ FLOP
- Yes
- Why it is tracked
- Training cost
- Record confidence
- Likely
Sources
Where this record came from and when it was last checked.
- Reference
- NVLM: Open Frontier-Class Multimodal LLMs
- Last updated
- 28 November 2025
What the numbers mean
Where it came from
NVLM-H 72B was published by NVIDIA, in the country recorded as United States of America, during October 2024. It comes out of an organisation categorised as industry.
It works in the domain of Vision, Language, and is recorded as performing the task of language modeling/generation, Vision-language generation, Question answering, Code generation, Translation, Quantitative reasoning.
Rather than being trained from scratch, it is derived from Qwen2-72B,InternViT-6B. That is the usual way a specialised model is produced.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
How it was trained
Training it took a computation budget of roughly 3 × 10²⁴ FLOP, on hardware recorded as NVIDIA H100 SXM5 80GB. That figure measures what producing the model cost, and has no bearing on how fast it answers.
Training consumed a corpus of around 125,829,120,000 tokens of text.
It is tracked in the underlying dataset for one reason in particular: training cost.
Answers
NVLM-H 72B — common questions
NVLM-H 72B— who created it?
It was published by NVIDIA, based in United States of America, an organisation categorised as industry.
NVLM-H 72B— when was it released?
It was published in October 2024. Capability per parameter has improved considerably since, so a newer model of the same size is often the better use of the same hardware.
NVLM-H 72B— what is it used for?
It works in the domain of Vision, Language, and is recorded as handling the task of language modeling/generation, Vision-language generation, Question answering, Code generation, Translation, Quantitative reasoning. These are the areas it was designed around; they describe intent rather than a hard boundary.
NVLM-H 72B— how much compute was used to train it?
Training consumed around 3 × 10²⁴ FLOP, on hardware recorded as NVIDIA H100 SXM5 80GB. That measures what producing the model cost and says nothing about how quickly it answers once trained — inference speed comes from memory bandwidth, not from the training budget.
NVLM-H 72B— what GPU do I need to run it?
None. This is a closed model — its weights were never published, so it cannot be downloaded or run on your own hardware at any price. It is reachable only through its provider.
NVLM-H 72B— is it open source?
No. Its weights have not been published, so it exists only as a service controlled by its owner.
NVLM-H 72B— how many parameters does it have?
It has a parameter count of 72B. 72B. That figure is the total, and it is what decides how much memory the model needs — roughly half a gigabyte per billion at the compression most people use.
The other direction
Looking at it from the other side?
This page starts from the model. If you already own a card and want to know everything it will run, start from the hardware instead.