Llama 4 Scout + ScaleRL
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
- Organisation type
- Industry
- Country
- United States of America
- Published
- 15 October 2025
- Authors
- Devvrit Khatri, Lovish Madaan, Rishabh Tiwari, Rachit Bansal, Sai Surya Duvvuri, Manzil Zaheer, Inderjit S. Dhillon, David Brandfonbrener, Rishabh Agarwal
What it does
The problem areas the model was built for. A model can carry several of each.
- Domain
- Language
- Task
- Language modeling/generation, Question answering, Code generation
- Base model
- Llama 4 Scout
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
- 109B
- Training data
- tokens
17Bx16 MoE (Scout) 109B total parameters
"7100 steps for Scout" SFT. "We run SFT using a batch size of 2M tokens, max sequence length of 12288 <..> on 32 H100 GPU nodes for approximately 4 epochs and 32B tokens in total." RL. "We allocate 14k generation budget during RL training, where 12k tokens are allocated to the intermediate reasoning (“thinking”), followed by 2k tokens for the final solution and answer. We sample 48 prompts in each batch, each with 16 generations per prompt. Thus, we get the total batch size as 768 completions …
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.
- How it was established
- Hardware
- Fine-tuning compute
- 1.4 × 10²³ FLOP
2500000000000000 FLOP/GPU/sec [GB200 reported] * 50000 GPU-hours * 3600 sec / hour * 0.3 [assumed utilization] = 1.35e+23 FLOP
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 GB200
- Chips used
- 80
- Chip-hours
- 50,000
- Power draw
- 187.5 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.
- Record confidence
- Confident
Sources
Where this record came from and when it was last checked.
- Reference
- The Art of Scaling Reinforcement Learning Compute for LLMs
- Last updated
- 28 November 2025
What the numbers mean
Where it came from
Llama 4 Scout + ScaleRL was published by Meta AI, in United States of America, in October 2025. industry is the category the publisher falls under.
It works in Language, and is recorded as doing language modeling/generation, Question answering, Code generation.
Its starting point was Llama 4 Scout — most models at this scale are adapted from an existing base rather than built from nothing.
Its weights were never published, so it can only be reached through its provider. No graphics card changes that.
Answers
Llama 4 Scout + ScaleRL — common questions
How many parameters does Llama 4 Scout + ScaleRL have?
Llama 4 Scout + ScaleRL has 109B parameters. 17Bx16 MoE (Scout) 109B total parameters. 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 Llama 4 Scout + ScaleRL?
Llama 4 Scout + ScaleRL was published by Meta AI, based in United States of America, categorised as industry.
When was Llama 4 Scout + ScaleRL released?
Llama 4 Scout + ScaleRL was published in October 2025.
What is Llama 4 Scout + ScaleRL used for?
Llama 4 Scout + ScaleRL works in Language, and is recorded as handling language modeling/generation, Question answering, Code generation. Models frequently carry more than one of each, and the tags describe purpose rather than capability limits.
What GPU do I need to run Llama 4 Scout + ScaleRL?
None. Llama 4 Scout + ScaleRL 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 Llama 4 Scout + ScaleRL open source?
No. Llama 4 Scout + ScaleRL has not had its weights published, so it exists only as a service controlled by its owner.
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.