What happened?
Google has announced its new model, Gemini 3.7 Flash. This model replaces Gemini 3.6 Flash, which launched only three weeks earlier. According to Google's Senior Director Tulsee Doshi, the new model resulted from core optimizations and developer feedback, delivering particular improvements in coding and agentic performance.
Doshi noted that the FrontierCode 1.1 Main test score rose from 34.4% to 43.6%, while the DeepSWE v1.1 score climbed from 49% to 65.3%. The WebDev Arena score also increased from 1,538 to 1,588. On GDP.pdf, which measures document processing performance, the score rose from 22% to 34%, and on AutomationBench, which measures workflow automation, it rose from 17% to 30.4%.
Why does it matter?
Such frequent model updates can be interpreted as Google's effort to maintain an image of continuous progress in AI. Throughout 2024 and 2025, Google moved quickly to catch up with rival labs, but that pace appears to have slowed in 2026. The flagship model Gemini 3.5 Pro, promised for a June launch at the I/O event in May, still hasn't been released.
The rapid succession of Flash models approaching version 4.0 has fueled speculation that Google is avoiding direct comparisons between 3.5 Pro and the latest models from OpenAI and Anthropic. Reports suggest Gemini's coding capabilities lag behind recent advances at other labs, and that Google is losing ground in AI capabilities.
How does pricing compare?
| Model | Input (1M tokens) | Output (1M tokens) |
|---|---|---|
| Gemini 3.7 Flash | $0.75 | $3.75 |
| Gemini 3.6 Flash | $1.50 | $7.50 |
| GPT 5.6 Luna (OpenAI) | $0.20 | $1.20 |
What's next?
- Gemini 3.7 Flash is available via the Gemini API, AI Studio, and Gemini Enterprise.
- The model currently only powers the Gemini Spark agent within the Gemini app, limited to AI Pro or Ultra subscribers.
- The standard chat interface still runs on Gemini 3.6 Flash.
- The release date for Gemini 3.5 Pro remains uncertain.