Google releases Gemini 3.5 Flash at Google I/O, claiming frontier-level agentic and coding performance at 4x the output speed of rival frontier models.
Key Takeaways
Outperforms Gemini 3.1 Pro on Terminal-Bench 2.1 (76.2%), GDPval-AA (1656 Elo), MCP Atlas (83.6%), and CharXiv Reasoning (84.2%).
Designed for long-horizon agentic workflows via the Antigravity harness, enabling parallel subagent execution for coding, document processing, and data tasks.
Default model for the Gemini app and AI Mode in Search; also available via Gemini API, AI Studio, Android Studio, and Gemini Enterprise Agent Platform.
Gemini Spark, a 24/7 personal AI agent powered by 3.5 Flash, begins rollout to trusted testers with Google AI Ultra beta planned for next week.
3.5 Pro is in internal use and expected to roll out next month.
Hacker News Comment Review
Pricing is the dominant concern: 3.5 Flash is priced at $1.50/$9.00 per million input/output tokens, a 3-5x jump over 2.5 Flash ($0.30/$2.50) and comparable to what 2.5 Pro cost, with commenters questioning whether capability gains justify the increase.
Quality-per-dollar skepticism is widespread; at least one benchmark comparison suggests 3.5 Flash underperforms 3.1 Pro on some dimensions while costing more per token overall.
A January 2025 knowledge cutoff on a model released May 2026 flagged as a significant gap, with commenters expressing concern about the 16-month lag.
Notable Comments
@reconnecting: Knowledge cutoff January 2025 against a May 2026 release date noted as alarming.
@GodelNumbering: Provides full pricing ladder from 2.5 Flash to 3.5 Flash, calling the jump unprecedented for a same-tier successive model.
