gemma-4-12B-it
Official · BF16 · safetensors
Runtime certified: Transformers
| Precision / quantization | Format | Bits | Publisher trust | Deployment state |
|---|---|---|---|---|
| BF16 | safetensors | 16 | Official | Runtime certified |
Gemma · General · Gemma 4
Gemma 4 12B IT: Specs, GPU requirements, hosting & deployment
Gemma 4 12B IT is a current General release in the Gemma family, certified from its immutable official source revision.
Total parameters12.0B
Active parametersDense model
Context window262,144 tokens
Release statusCurrent · Current
ArchitectureGemma4UnifiedForConditionalGeneration
ModalitiesText, Image, Audio
LicenseApache License 2.0
Deployment stateDeployable with certified configuration
Commercial stateCommercial Use Allowed
Release date2026-05-23
Architecture, strengths and use cases
Gemma 4 12B IT is an official self-hostable Google release in the Gemma General track. It uses Gemma4UnifiedForConditionalGeneration, carries 12B, supports a verified context window of 262,144 tokens, and exposes only source-certified checkpoints and runtime combinations. Its release, license and deployment facts are tied to the official repository revision rather than inferred from the family name.
Gemma 4 12B IT uses Gemma4UnifiedForConditionalGeneration with 12B. The stored architecture profile records 48 layers, hidden width 3840, 16 attention heads and a 262,144 token context ceiling where the official configuration publishes those fields.
Verified Gemma4UnifiedForConditionalGeneration architecture
262,144 token certified context
Official BF16, Unknown checkpoint coverage
Private LLM inference
Agentic workflows
Long-context document processing
Vision-language analysis
Audio understanding
Certified checkpoints and quantizations
Only exact, source-certified variants attached to this release are shown.
Recommended hardware
Use only the listed transformers runtime configuration with the exact checkpoint precision. The recommendation engine checks weight memory, KV cache, runtime overhead, safety headroom, GPU architecture and topology before ranking rentable and reference hardware.
Estimated weight memory23.02 GB
KV cache3 GB
Runtime overhead9.4 GB
Total with safety margin43.93 GB
Recommended best value
1× NVIDIA H200 SXM 141GB
141 GB total VRAM · nvswitch
97.07 GB headroom · Limited capacity
USD 2,345 / monthLowest-cost viable
2× NVIDIA L4 24GB
48 GB total VRAM · pcie
4.07 GB headroom · Available
USD 500 / monthHighest performance
1× NVIDIA H200 SXM 141GB
141 GB total VRAM · nvswitch
97.07 GB headroom · Limited capacity
USD 2,345 / monthLicense and self-hosting considerations
The full 12B resident weight set must fit across the selected GPUs. Context and concurrency add KV-cache memory; MoE active parameters do not replace resident-weight memory.
Gemma 4 12B IT is published under Apache License 2.0. The release-scoped official license link and verification date are retained with the catalogue record.
Commercial Use Allowed. The official license permits commercial use; its obligations still apply.
Alternatives and sibling variants
Gemma 4 31B IT, Gemma 4 26B-A4B IT, Gemma 3 27B IT
Release timeline
google/gemma/gemma-4-12b-it
Source details
Source revision: c0b5ed3f8356289e6ba070b6a9006fc66e475c195c78a18218f84c13055d5e67
Frequently asked questions
How much VRAM does Gemma 4 12B IT require?
The exact estimate depends on checkpoint precision, context and concurrency; the page recommendation uses the certified memory engine and compatible runtime configurations.
Which runtimes support Gemma 4 12B IT?
Certified upstream runtime evidence is available for transformers.
Can Gemma 4 12B IT be used commercially?
The stored commercial classification is Commercial Use Allowed. Consult the official Apache License 2.0 terms for the final obligations.
Official references
Last verified: 2026-08-26