Google's budget model is now two iterations ahead of its premium version.

      Google has launched Gemini 3.7 Flash, showcasing significant improvements in coding benchmarks and an introductory price of $0.75 per million input tokens. Meanwhile, Gemini 3.5 Pro continues to lag behind schedule by several months, and Google has not confirmed its release status.

      With the release of Gemini 3.7 Flash, there is still uncertainty regarding the arrival of its flagship model. The new workhorse model displays substantial advancements in coding, an area where Google has been facing challenges. The model it is intended to support, Gemini 3.5 Pro, is still months overdue.

      The performance metrics are notable. On FrontierCode 1.1, the new model achieves a score of 43.6%, compared to 34.4% for the previous version; on DeepSWE v1.1, it reaches 65.3% against 49.0%; and on AutomationBench, it more than doubles from 17.0% to 30.4%.

      Google aims to encourage usage with its pricing, charging $0.75 per million tokens for input and $3.75 for output until the end of December, after which both rates will increase. This model also powers Gemini Spark for AI Pro and Ultra subscribers in over 160 countries.

      Regarding safety, the updates are focused on frontier categories. Google has implemented protections against cyber threats as well as chemical, biological, radiological, and nuclear misuse, aligning with its Frontier Safety commitments.

      The release rhythm is noteworthy, occurring just three weeks after the previous Flash model, with three Flash models released by Google in July, including one tailored for security.

      Conversely, the Pro line has seen no advancement during this period. The upcoming Pro model is still several months behind due to its coding performance not meeting internal expectations, which is unfortunate considering that coding is a lucrative area.

      There’s a possibility that it might not be released at all. Axios has reported that Google might bypass 3.5 Pro entirely and proceed directly to Gemini 4 Pro, and the company has refrained from commenting on the future of this model.

      In July, Sundar Pichai informed investors that Google plans to accelerate the release of models and is already investing significantly in computing resources to train Gemini 4. Releasing the less expensive tier every three weeks offers a perception of speed while the pricier model experiences delays.

      The internal impact of these delays is significant, affecting morale at DeepMind during a time when rival labs are actively hiring.

       Published August 13, 2026 - 8:09 pm UTC

       Back to top