Google makes Gemini 3.7 Flash the default model for Antigravity agents

Google's Gemini API documentation says Gemini 3.7 Flash is now the default model for the Antigravity agent in Gemini Managed Agents and the Google Antigravity SDK. That is the concrete change: a model selection that affects a managed-agent path, rather than a claim about the model's release state. Google separately describes Antigravity Teamwork as a framework in which autonomous agents collaborate, critique, and iterate over hours or days on long-horizon work.

The API specification identifies the endpoint as `gemini-3.7-flash`, with text, image, video, audio, and PDF inputs, text output, a 1,048,576-token input limit, and a 65,536-token output limit. It documents support for caching, code execution, file search, function calling, search grounding, structured outputs, URL context, and configurable thinking levels. The same page lists computer use as preview-only and says audio generation, image generation, and the Live API are not supported, so an Antigravity default does not turn the model into a universal multimodal endpoint.

Google's migration guide positions 3.7 Flash as the target for workloads using Gemini 3.6 Flash, Gemini 3.5 Flash, Gemini 3 Flash Preview, or Gemini 3.1 Pro. The transition is not only a model-string replacement: Google instructs users to remove deprecated `temperature`, `top_p`, and `top_k` parameters, replace `thinking_budget` with `thinking_level`, remove `candidate_count`, and standardize multi-turn flows on `previous_interaction_id`. That makes the comparison with earlier Flash generations practical as well as performance-oriented.

Google attributes the default change to improved performance and reasoning, and its Antigravity post reports research, systems, and open-source outcomes from pairing 3.7 Flash with Teamwork. Those accounts include a reported 71% on TCSBench, a RISC-V simulator result, and upstream optimization claims, but they are published by Google rather than a neutral evaluator. They are useful evidence of the workflows Google is targeting, not a substitute for an independently reproduced comparison on a team's own codebase and tool chain.

For technical users, the immediate implication is to audit agent configurations and integration tests before assuming an unchanged runtime. Google says the underlying Antigravity model remains configurable through `agent_config`, while its thinking guidance frames low effort as latency-oriented, medium as the default, and high as a higher-cost setting for difficult reasoning and tool use. The source set establishes Google's documented controls and positioning; it does not independently establish that every existing agent will become more reliable or less expensive after the default changes.