GTA1 7B 2507 is a multimodal model from Salesforce that accepts both text and image inputs, operating within a large 128K token context window. At 7 billion parameters, it sits in the mid-size range — capable enough for substantive tasks while remaining accessible for local deployment under an Apache 2.0 license. Specific behavioral characteristics and capability details are limited beyond its vision-language architecture.