Best larger-system route
Frontier Native 2.3TB
The smallest system in the range that reaches this model's preferred working allowance.
Explore Frontier Native 2.3TBMeta model
Llama 4 Scout Hardware Requirements & Compatibility
The official BF16 checkpoint contains about 217.3GB of weights. A sharded 256GB server is the tight arithmetic floor, while 384GB provides a more credible operating envelope.
- Model version
meta-llama/Llama-4-Scout-17B-16E-Instruct- Family and variant
- Llama 4 Scout · Llama-4-Scout-17B-16E-Instruct
- Source version
92f3b1597a19- Source updated
- 22 May 2025
Meta Llama 4 Scout
meta-llama/Llama-4-Scout-17B-16E-Instruct - Minimum GPU memory
- At least 240GB aggregate for BF16 weights and minimal overhead
- Recommended hardware
- 384GB aggregate for a less constrained sharded deployment
- Licence
- Llama 4 Community Licence
- Useful for
- Language & reasoning, Vision & OCR
Buyer verdict
Where Llama 4 Scout is a sensible fit
Llama 4 Scout is relevant where the team wants the Llama ecosystem, multimodal input and long-context experiments, and can accept Meta's gated licence and multi-GPU deployment needs.
These figures apply to the named model version. Quantisation, fine-tuning, context length, image resolution, batch size and serving software can materially change the hardware needed.
Hardware requirements
- Minimum
- At least 240GB aggregate for BF16 weights and minimal overhead
- Recommended
- 384GB aggregate for a less constrained sharded deployment
A quantised build can change this floor, but it is a separate artefact.
Long contexts and images can require substantially more memory.
Technical specification
Llama 4 Scout model and hardware facts
Specifications shown for source version
92f3b1597a195b523d8d9e5700e57e4fbb8f20d3, updated
22 May 2025.
- Architecture
- MoE, 109B total / 17B active
- Modalities
- Text + image → text
- Repository weights
- 217.28GB BF16
- Access
- Gated model
- Licence
- Llama 4 Community Licence
- Serving
- Runtime-specific multimodal support required
Product compatibility
Llama 4 Scout compatibility across all 11 GPU systems
- Systems that fit without model splitting
- 2
- Systems needing multi-GPU validation
- 4
A system is listed as fitting when its GPU memory meets the requirement shown above. Speed, usable context, batch size and concurrent users still need testing with the final model and software configuration.
sme workstation
Team 32
- Per GPU
- 32GB
- Total GPU memory
- 32GB
- GPU count
- 1
This system provides 32GB per GPU and 32GB in total. The model requires 240GB aggregate across at least 4 GPUs.
sme workstation
Company 64
- Per GPU
- 32GB
- Total GPU memory
- 64GB
- GPU count
- 2
This system provides 32GB per GPU and 64GB in total. The model requires 240GB aggregate across at least 4 GPUs.
sme workstation
Studio 96
- Per GPU
- 96GB
- Total GPU memory
- 96GB
- GPU count
- 1
This system provides 96GB per GPU and 96GB in total. The model requires 240GB aggregate across at least 4 GPUs.
sme workstation
Studio 192
- Per GPU
- 96GB
- Total GPU memory
- 192GB
- GPU count
- 2
This system provides 96GB per GPU and 192GB in total. The model requires 240GB aggregate across at least 4 GPUs.
pcie rack
Value Rack 128
- Per GPU
- 32GB
- Total GPU memory
- 128GB
- GPU count
- 4
This system provides 32GB per GPU and 128GB in total. The model requires 240GB aggregate across at least 4 GPUs.
pcie rack
Value Rack 256
- Per GPU
- 32GB
- Total GPU memory
- 256GB
- GPU count
- 8
Total installed memory is sufficient, but the model must be divided across GPUs. Confirm that the serving software supports this GPU layout and test the required context and speed.
pcie rack
Enterprise 384
- Per GPU
- 96GB
- Total GPU memory
- 384GB
- GPU count
- 4
Total installed memory is sufficient, but the model must be divided across GPUs. Confirm that the serving software supports this GPU layout and test the required context and speed.
pcie rack
Enterprise 768
- Per GPU
- 96GB
- Total GPU memory
- 768GB
- GPU count
- 8
Total installed memory is sufficient, but the model must be divided across GPUs. Confirm that the serving software supports this GPU layout and test the required context and speed.
pcie rack
H200 1.1TB
- Per GPU
- 141GB
- Total GPU memory
- 1,128GB
- GPU count
- 8
Total installed memory is sufficient, but the model must be divided across GPUs. Confirm that the serving software supports this GPU layout and test the required context and speed.
frontier partner
Frontier Native 2.3TB
- Per GPU
- 288GB
- Total GPU memory
- 2,304GB
- GPU count
- 8
Also meets this model page's recommended working allowance.
Available GPU memory exceeds the calculated or estimated requirement. Confirm the final precision, software, workload and performance before purchase.
frontier partner
Frontier Rack 20TB
- Per GPU
- 288GB
- Total GPU memory
- 20,736GB
- GPU count
- 72
Also meets this model page's recommended working allowance.
Available GPU memory exceeds the calculated or estimated requirement. Confirm the final precision, software, workload and performance before purchase.
Deployment reality
Strengths, limits and runtime route
What it is good at
- Large Llama ecosystem and broad serving-framework interest.
- Text and image input with a sparse 16-expert architecture.
- 17B active parameters despite a 109B total checkpoint.
- Suitable for multimodal retrieval and long-document evaluation when the context is carefully bounded.
Where to be cautious
- Access is gated and governed by the Llama 4 community licence.
- The official BF16 weights are substantially larger than one 96GB GPU.
- Advertised context ceilings do not establish practical cache, latency or quality on a given server.
- Commercial terms, acceptable-use policy and attribution need review before deployment.
Serving software
- vLLM after version check
- Transformers
- SGLang after version check
Before installation
- Accept Meta's terms through the intended legal entity before downloading the checkpoint.
- Use an exact runtime version with documented Llama 4 multimodal support.
- Begin with one image and bounded context in acceptance tests.
- Do not substitute a community quantisation without recording its publisher, revision and licence inheritance.
System requirements
- GPU layout
- Plan for at least 4 GPUs and validate model parallelism on the final topology.
- Context and cache
- The model context ceiling is not a guaranteed serving target. KV cache, batch size and concurrent sessions need separate capacity tests.
- System RAM
- Size system memory for model loading, runtime overhead, preprocessing and any CPU offload used by the final configuration.
- Storage
- Allow space for the pinned checkpoint, runtime images, caches, logs and at least one rollback version.
- Serving software
- Validate the exact checkpoint with vLLM after version check, Transformers, SGLang after version check before acceptance.
- Representative workload
- Benchmark representative prompts or media at the required context, quality, latency and concurrency.
Performance: speed, usable context and concurrency depend on the selected system, software and workload. No benchmark is quoted on this page.
Commercial and legal boundary
Llama 4 Community Licence
Commercial use: conditional
Commercial use is subject to Meta's model-specific terms, acceptable-use policy and access approval.
Always retain the applicable notices and recheck the live terms for the intended organisation, territory, use and distribution route. Obtain legal advice where required; the official licence governs use.
Read the official licenceTechnical questions
Llama 4 Scout deployment FAQ
Can Llama 4 Scout run on 96GB?
Not as the named BF16 checkpoint. Community quantisations may reduce memory, but the exact conversion and runtime need a separate record.
Is Value Rack 256 compatible?
Its aggregate total clears a tight arithmetic floor, but the memory is split across eight PCIe GPUs. Treat it as conditional until the exact runtime and topology pass.
Is the Llama licence open source?
It is an open-weight community licence with its own conditions, not Apache 2.0 or MIT. Review the current terms for the intended use.
What a complete Llama 4 Scout deployment needs
GPU memory is only one part of the system. Storage, data access, serving software, monitoring and administrator handover also affect a reliable deployment.
