DevOps / SRE / Platform · 13.08.2026, 20:10 UTC
The AI model that just scored 65% on DeepSWE isn’t the one Google promised.
| Schweregrad | info |
|---|---|
| Kategorie | DevOps / SRE / Platform |
| Quelle | The New Stack ↗ |
| Veröffentlicht | 13.08.2026 UTC |
Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.
Google has a new Gemini model, and no, it is not Gemini 3.5 Pro. Gemini 3.7 Flash launched Thursday as the company’s new workhorse for coding agents and automated business workflows, just three weeks after Gemini 3.6 Flash became available.
The quick turnaround may cause some version fatigue, but Google says the new model is better at writing code and handling the longer workflows that keep agents on track. Google has paired those improvements with an introductory API price that is half the original cost of Gemini 3.6 Flash, but with a caveat; that price will increase on January 1, 2027.
Google says Gemini 3.7 Flash is less likely to get stuck when something goes wrong and better at knowing when it needs more information before moving forward.
Coding benchmarks jump sharply
Google says Gemini 3.7 Flash is less likely to get stuck when something goes wrong and better at knowing when it needs more information before moving forward. The model scored 43.6% on FrontierCode 1.1 Main, up from 34.4% for Gemini 3.6 Flash. Its DeepSWE v1.1 score climbed from 49% to 65.3%. The pattern echoes what happened when DeepSeek’s smaller model outperformed its own flagship — leaner models are consistently punching above their weight class. The WebDev Arena score rose more modestly, from 1,538 to 1,588.
But benchmarks cannot show how reliably that will happen inside a company’s own environment. Google seems focused on helping agents work through a codebase without getting stuck. If Gemini 3.7 Flash can recover from a bad first attempt and reach the right result faster, it could save …