Google releases Gemini 3.5 Flash at Google I/O, claiming frontier agentic and coding performance at 4x the output token speed of rival models.
Key Takeaways
3.5 Flash tops Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), MCP Atlas (83.6%), and CharXiv Reasoning (84.2%), outperforming Gemini 3.1 Pro on agentic benchmarks.
Priced at $1.50/M input and $9.00/M output tokens; Google claims under half the cost of other frontier models for agentic workloads.
Available now via Google Antigravity, Gemini API, AI Studio, Android Studio, Gemini Enterprise, and the Gemini app globally.
Gemini Spark, a 24/7 personal AI agent powered by 3.5 Flash, is rolling out to trusted testers with Google AI Ultra beta planned for next week.
Enterprise partners including Shopify, Macquarie Bank, Salesforce, Ramp, Xero, and Databricks are piloting agentic workflows with the model.
Hacker News Comment Review
The price jump is the dominant concern: 3.5 Flash costs $1.50/$9.00 per million tokens, a 3x increase over 3.1 Flash and roughly equal to what 2.5 Pro cost, breaking the usual Flash value proposition.
Commenters on Artificial Analysis data argue the model is less capable per token than the price implies, with GPT-5.5 still seen as stronger on benchmarks despite Google’s claims.
Skepticism runs high about Gemini 3.0 being skipped to preview-only status and whether Google can close the vibes gap on tool calling before users commit to competing platforms.
Notable Comments
@GodelNumbering: Lays out the full pricing ladder from 2.5 Flash to 3.5 Flash, noting the unprecedented 3x jump and parity with 2.5 Pro pricing.
@asar: Flags that 3.5 Flash input price is 6x the 3.1 Flash Lite rate, a sharp context anchor for cost-sensitive builders.
