Gemini 3.7 Flash Doubles Coding Performance Over 3.6 Flash at Half the Price
Google released Gemini 3.7 Flash on August 13, 2026 — arriving just three weeks after Gemini 3.6 Flash and delivering what Google describes as its most significant algorithmic improvement in the Flash tier to date. The model is live now across the Gemini API, AI Studio, Vertex AI, and the Gemini Enterprise Agent Platform.
The headline claim is a near-doubling of performance on long-horizon coding tasks. On DeepSWE, the benchmark that measures autonomous software engineering on real repositories, Gemini 3.7 Flash scored 65.3% versus 49.0% for its predecessor — a 16-point gain. On AutomationBench, which tests multi-step agentic task completion, 3.7 Flash reached 30.4% against 3.6 Flash's 17.0%. Google attributes the gains to enhanced post-training on agentic trajectories and reinforcement learning focused on coding and multi-step execution, not to an increased model size or expanded context window — both models share the same 1-million-token context and 65,536 output token limit.
Pricing at launch is $0.75 per million input tokens and $3.75 per million output tokens — exactly half the standard rates Google charged for 3.6 Flash — with this introductory rate locked in through December 31, 2026. The model includes configurable thinking effort levels (low, medium, high) that let callers trade latency for reasoning depth, a control surface that first appeared in the higher-tier Gemini 3.1 Pro family.
The 3.7 Flash release keeps up a pace of rapid iteration in Google's Flash tier that has been consistent through 2026. For developers already using Gemini Managed Agents — which gained async background execution and native MCP server support in July — the performance jump on agentic tasks is the most immediately actionable improvement in this release.
