AI News

Gemini 3.5 Flash Is Here, Google’s Strongest Agentic & Coding Model Yet

Gemini 3.5 Flash

Google has officially announced Gemini 3.5 Flash at today’s I/O 2026 event. The company claims the new model is its fastest and most capable Flash model yet. According to Google, Gemini 3.5 Flash offers frontier-level reasoning and top-tier coding performance while maintaining the ultra-fast response times the Flash lineup is known for. The company also claims the model can operate at roughly four times the speed of comparable flagship systems, often while costing less than half as much. Notably, Gemini 3.5 Flash is available starting today.

Gemini 3.5 Flash is chasing the biggest rivals directly

With rivals like OpenAI, Anthropic, and Google is becoming increasingly competent, Gemini 3.5 Flash appears to be designed to prove that speed no longer has to come at the expense of intelligence. To prove model’s worth, Google shared several benchmark comparisons during the announcement, and the numbers are surprisingly pretty solid.

In Terminal-Bench 2.1, Gemini 3.5 Flash reportedly scored 76.2% and outperforms Gemini 3.1 Pro at 70.3% and Gemini 3 Flash at 58%. The company also talkeed about the models’ solid agentic workflow performance. On GDPval-AA Elo, Gemini 3.5 Flash grabbed a score of 1656, while Gemini 3.1 Pro landed at 1314.

Meanwhile, on MCP Atlas scaled tool usage, the new model scored 83.6%, once again outpacing the previous Gemini versions. Not to forget, Google has also compared the new model directly against rivals like GPT-5.5 and Claude Opus 4.7. While GPT-5.5 still appears slightly stronger in raw coding and abstract reasoning tests, Gemini 3.5 Flash reportedly performs competitively across most categories.

Gemini 3.5 Flash Benchmark
Image: Google

Google may have just changed the speed race

Output speed is where things become especially interesting. Based on third-party Artificial Analysis testing from May 13, Gemini 3.5 Flash reportedly generates around 289 tokens per second. That figure is dramatically faster than Claude Opus 4.7 at 67 tokens per second and GPT-5.5 at 71. Even Gemini 3.1 Pro reportedly reaches only 135.

With Google now pushing both flagship-level reasoning and near real-time responsiveness simultaneously, Gemini 3.5 Flash could end up becoming one of the company’s biggest developer-focused launches yet.

Rishaj Upadhyay
Rishaj is a tech journalist with a passion for AI, Android, Windows, and all things tech. He enjoys breaking down complex topics into stories readers can relate to. When he's not breaking the keyboard, you can find him on his favorite subreddits, or listening to music/podcasts
You may also like
More in:AI News