Apollo 7B
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
- Meta AI,Stanford University
- Organisation type
- Industry,Academia
- Country
- United States of America
- Published
- 13 December 2024
- Authors
- Orr Zohar, Xiaohan Wang, Yann Dubois, Nikhil Mehta, Tong Xiao, Philippe Hansen-Estruch, Licheng Yu, Xiaofang Wang, Felix Juefei-Xu, Ning Zhang, Serena Yeung-Levy, Xide Xia
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Video, Language, Multimodal
- Task
- Video description
- Base model
- Qwen2.5-7B,SigLIP 400M
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
- 7B
- Training data
- tokens
"We employed the Qwen2.5 (Yang et al., 2024) series of Large Language Models (LLMs) at varying scales to serve as the backbone for Apollo. Specifically, we utilized models with 1.5B, 3B, and 7B parameters. Following our analysis in Sec. 4, we used a SigLIP-SO400M (Zhai et al., 2023) encoder combined with an InternVideo2 (Wang et al., 2024d) video encoder" there is a confusion with Qwen2.5 link - it points to Qwen 2 paper instead of Qwen 2.5 release notes
1. Alignment: In this phase, we trained on a 198K mixture of 50/50 image and video captions. 2. Vision Pretraining: We tuned the encoders using a video-only caption dataset of 396K samples. 3. Supervised Fine-tuning (SFT): We trained on a mixture of text, image, multi-image, and video data, with a total of 3.2 million samples. 3200000 videos *36%*~40 sec*~128 tokens / sec = 5898240000 video tokens during SFT assuming about the same amount of image and text tokens. total size of the dataset was…
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 A100
- Chips used
- 128
- Power draw
- 100.7 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.
- Why it is tracked
- SOTA improvement
- Record confidence
- Confident
Apollo-7B is state-of-the-art compared to 7B LMMs with a 70.9 on MLVU, and 63.3 on Video-MME
Sources
Where this record came from and when it was last checked.
- Reference
- Apollo: An Exploration of Video Understanding in Large Multimodal Models
- Last updated
- 28 November 2025
What the numbers mean
Where it came from
Apollo 7B was published by Meta AI,Stanford University, in United States of America, in December 2024. The organisation is categorised as industry,Academia.
It works in Video, Language, Multimodal, and is recorded as doing video description.
It builds on Qwen2.5-7B,SigLIP 400M, which is why it shares that model's general shape and size.
Because the weights are not available, none of the hardware figures elsewhere on this site apply to it.
Training and provenance
The reason it appears in this catalogue at all is sOTA improvement.
Answers
Apollo 7B — common questions
When was Apollo 7B released?
Apollo 7B was published in December 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.
What is Apollo 7B used for?
Apollo 7B works in Video, Language, Multimodal, and is recorded as handling video description. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
What GPU do I need to run Apollo 7B?
None. Apollo 7B 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.
Is Apollo 7B open source?
No. Apollo 7B has not had its weights published, so it exists only as a service controlled by its owner.
How many parameters does Apollo 7B have?
Apollo 7B has 7B parameters. "We employed the Qwen2.5 (Yang et al., 2024) series of Large Language Models (LLMs) at varying scales to serve as the backbone for Apollo. Specifically, we utilized models with 1.5B, 3B, and 7B parameters. Following our analysis in Sec. 4, we used a SigLIP-SO400M (Zhai et al., 2023) encoder combined with an InternVideo2 (Wang et al., 2024d) video encoder" there is a confusion with Qwen2.5 link - it points to Qwen 2 paper instead of Qwen 2.5 release notes. 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.
Who created Apollo 7B?
Apollo 7B was published by Meta AI,Stanford University, based in United States of America, categorised as industry,Academia.
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.