Gemma 4 E4B
By Google DeepMind
Chat, reasoning and image understanding.
Available to invited pilot accounts on our scheduled Cape Town host. Prices below are proposals; Billing in your workspace shows the active configured rates.
Model capabilities
Text, Reasoning, Vision. Check the service limits below for the supported pilot subset.
Read the official model cardPublished context128K
Parameters8B
Size bandLight
Publisher: 4.5B effective parameters, 8B including embeddings. The pilot serves a quantized model with a smaller context and one PNG input; audio and tools are not enabled.
Published benchmarks
No verified benchmark scores are listed for this candidate yet.
Publisher-reported results, not Embiro tests. Scores across different suites or evaluation setups should not be compared.
Pilot service limits
One inline PNG, up to 1024×1024; output 640–2048 tokens including reasoning. No tools.
API model ID: gemma-4-e4b