Gemma 3 4B
Gemma 3 4B It, built by Google, is a multimodal chat model that handles text and image input and generates text output, with capabilities in vision tasks such as question answering and image understanding. It is notable for its large context window of 32,768 tokens, allowing it to process lengthy inputs.
Input
Output
Context
131K
Max Output
8K
Parameters
4.3B
Technical Specifications
Input Modalities
Output Modalities
Capabilities
Loading capabilities…
Benchmarks
Estimated GPU requirements for Gemma 3 4B
Estimates based on INT8 quantization at up to 32K context. A count above one assumes tensor parallelism across the cards. Actual requirements vary by framework and configuration.
More models by Google
The creator's other models in the catalog, with their context, size and license where published.
Compare it, or keep browsing
Put this model beside its alternatives on the same evidence, or go back to the full catalog.
Frequently asked questions about Gemma 3 4B
Answered from the entry's own fields: context, license, modalities, evidence, serving and the memory to self-host.
Gemma 3 4B advertises a context window of 131,072 tokens, with a maximum output of 8,192 tokens in a single response. The figure is the creator's published maximum; a given host may serve less, and the gateway routes on what each host actually serves.
Yes. Gemma 3 4B is an open-weight model released under the Gemma license, so the weights can be downloaded and self-hosted within that license's terms.
Gemma 3 4B accepts Text and Image and produces Text. The capabilities card on this page lists which API features each deployment honours, such as function calling and structured output, with the source each was checked against.
Gemma 3 4B has published results from Artificial Analysis, shown by suite in the benchmarks card above exactly as the publisher reported them. Scores are not combined across suites, and a suite that has not measured Gemma 3 4B is shown as not available rather than estimated.
Yes. Gemma 3 4B is served on the managed pool through the OpenAI-compatible endpoint as google/gemma-3-4b-it, pinned by name or chosen by routing when it is the best fit for a request. The Try in Playground button opens it directly.
About 4.6 GB at INT8 for the weights and a default context, from the catalog's 4.3B parameter count; FP16 needs roughly twice that, and long contexts or many concurrent requests add KV cache on top. The GPU section on this page lists cards that hold it, and the capacity planner sizes it for your context length and traffic.
Fields collected from public registries, host APIs and benchmark publishers, each tagged with its source.
Last updated: Sep 13, 2026
Put intelligence in the middle.
One gateway in front of every model, with your policies applied and every decision on record. Start with $5 of credit and 5,000 routing decisions a month, no card required.