Google AI positioned Gemini 3.7 Flash as a lower-priced model for coding and multi-step AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → work in its August 13, 2026 announcement. It reported a DeepSWE v1.1 software-engineering score of 65.3%, compared with 49.0% for 3.6 Flash, alongside introductory input and output prices at half the older model's original rates.
The announcement also describes greater effort on planning and tool calls. That connects the release to workflows where an assistant must carry out several steps, inspect results and adapt. The reported benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → gain supports evaluating that capability; it does not establish how much oversight a particular production workflow can remove.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
