Best larger-system route
Studio 96
The smallest system in the range that reaches this model's preferred working allowance.
Explore Studio 96Google model
Gemma 3 27B Hardware Requirements & Compatibility
The BF16 repository weights total about 54.9GB. A 64GB route is the tight file-size floor, while one 96GB GPU is the clean product fit. Quantisation-aware variants can reduce the requirement but must be named separately.
- Model version
google/gemma-3-27b-it- Family and variant
- Gemma 3 27B · gemma-3-27b-it
- Source version
005ad3404e59- Source updated
- 21 March 2025
Google Gemma 3 27B
google/gemma-3-27b-it - Minimum GPU memory
- 64GB GPU for BF16 at a bounded context
- Recommended hardware
- One 96GB GPU
- Licence
- Gemma Terms of Use
- Useful for
- Language & reasoning, Vision & OCR
Buyer verdict
Where Gemma 3 27B is a sensible fit
Gemma 3 27B is a practical high-quality multimodal model for document, image and general assistant work where a single professional GPU and Google's gated Gemma terms are acceptable.
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
- 64GB GPU for BF16 at a bounded context
- Recommended
- One 96GB GPU
The 54.9GB weight total leaves modest runtime headroom.
This avoids model sharding and leaves a more useful multimodal operating envelope.
Technical specification
Gemma 3 27B model and hardware facts
Specifications shown for source version
005ad3404e59d6023443cb575daa05336842228a, updated
21 March 2025.
- Parameters
- 27B
- Context
- 128K tokens
- Modalities
- Text + image → text
- Repository weights
- 54.86GB BF16
- Access
- Gated licence acceptance
- Licence
- Gemma terms
Product compatibility
Gemma 3 27B compatibility across all 11 GPU systems
- Systems that fit without model splitting
- 7
- Systems needing multi-GPU validation
- 0
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 64GB on one GPU.
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 64GB on one GPU.
sme workstation
Studio 96
- Per GPU
- 96GB
- Total GPU memory
- 96GB
- GPU count
- 1
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.
sme workstation
Studio 192
- Per GPU
- 96GB
- Total GPU memory
- 192GB
- GPU count
- 2
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.
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 64GB on one GPU.
pcie rack
Value Rack 256
- Per GPU
- 32GB
- Total GPU memory
- 256GB
- GPU count
- 8
This system provides 32GB per GPU and 256GB in total. The model requires 64GB on one GPU.
pcie rack
Enterprise 384
- Per GPU
- 96GB
- Total GPU memory
- 384GB
- GPU count
- 4
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.
pcie rack
Enterprise 768
- Per GPU
- 96GB
- Total GPU memory
- 768GB
- 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.
pcie rack
H200 1.1TB
- Per GPU
- 141GB
- Total GPU memory
- 1,128GB
- 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 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
- Text and image input in a model small enough for one 96GB GPU.
- 128K context and multilingual coverage.
- Google provides quantisation-aware variants for smaller deployments.
- Broad Transformers and serving-library support.
Where to be cautious
- Hugging Face access requires acceptance of the Gemma licence.
- The BF16 weights do not fit one 32GB GPU.
- Image resolution, count and context must be represented in memory testing.
- Licence terms are model-specific rather than Apache or MIT.
Serving software
- Transformers
- vLLM after version check
- Gemma reference tooling
Before installation
- Use the exact 27B instruction checkpoint and pin the accepted licence state.
- For 32GB hardware, select an official or attributable quantised variant and record quality trade-offs.
- Test OCR-like tasks with the actual document resolution and layout.
- Reserve memory for the vision encoder, cache and runtime.
System requirements
- GPU layout
- The minimum can fit on one GPU; multiple GPUs may still be useful for replicas or throughput.
- 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 Transformers, vLLM after version check, Gemma reference tooling 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
Gemma Terms of Use
Commercial use: conditional
Use is subject to Google's current Gemma terms and prohibited-use policy.
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
Gemma 3 27B deployment FAQ
Can Gemma 3 27B run on 32GB?
Not as the named BF16 checkpoint. An attributable quantised variant may fit, but it needs its own precision, source and context record.
Which product is the simplest fit?
Studio 96 is the first straightforward single-GPU fit for the official BF16 weights.
Can Gemma 3 process documents?
It can analyse document images and text, but OCR accuracy, page resolution and retrieval design should be tested on representative files.
What a complete Gemma 3 27B 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.
