Google launched Gemini 3.5 Flash at I/O 2026 on May 19, positioning the model as a Flash-tier release that beats the prior Gemini 3.1 Pro flagship on coding and agentic benchmarks while undercutting it on price. The headline framing is agents, not chatbots.

Try It Today

Terminal window with API key card for Gemini 3.5 Flash access

Three surfaces are live now. Developers can hit Gemini 3.5 Flash through the Gemini API and the AI Studio playground for prompt-level testing. Creators on Workspace get it inside the Gemini app and AI Mode in Google Search at no extra cost. Pricing is $1.50 per 1M input tokens and $9.00 per 1M output tokens, with cached input at $0.15 per 1M, which is 40% under Gemini 3.1 Pro across input and output (per llm-stats coverage). First-cycle workflow today: swap your Gemini 3.1 Pro endpoint for 3.5 Flash, re-run your top coding prompt, and compare output speed. The price cut alone makes the test worth doing.

Why It Matters

40 percent cost reduction with lightning bolt for Gemini 3.5 Flash

Google's framing reads as a category bet. The Flash tier has historically been the cheap, fast, lower-quality option below Pro. With 3.5 Flash beating 3.1 Pro on every coding and agent benchmark per TechCrunch's coverage, the tier hierarchy inverts mid-cycle. Creators using Gemini for code generation, agent tooling, or long-horizon task automation get a faster and cheaper model that is also more capable than what they were paying double for last month. The release sits alongside the rest of the I/O 2026 Gemini push.

Key Details

Benchmark comparison bars showing Gemini 3.5 Flash performance

Benchmark spread per llm-stats:

  • Terminal-Bench 2.1: 76.2% (3.1 Pro 70.3%)
  • MCP Atlas: 83.6% (3.1 Pro 78.2%)
  • CharXiv Reasoning: 84.2% (3.1 Pro 83.3%)
  • GDPval-AA Elo: 1656 (3.1 Pro 1314)

Google claims the model runs four times faster than comparable frontier models measured in output tokens per second. The Pro variant is in internal use and is expected to roll out next month. The drop also lands across coding-agent platforms that route through the Gemini API.

What to Do Next

Coding workflows that currently call Gemini 3.1 Pro should swap to 3.5 Flash this week. Agent workflows that batched at the Pro tier for capability reasons can drop down to Flash. Run a benchmark sweep on your existing prompts to see whether the academic reasoning trail-off matters for your use case. The Pro variant arrives next month if you need to hedge.