Google says Gemini 3.5 Flash rivals 'large flagship models' for coding and agentic tasks
Source: Engadget
Overview
Google has unveiled Gemini 3.5, beginning with the Gemini 3.5 Flash model. Google claims Flash delivers “frontier‑level intelligence at exceptional speed,” allowing users to avoid the usual trade‑off between quality and latency. Announced at Google I/O 2026, Gemini 3.5 Flash will become Google’s default AI model (distinct from the earlier Flash‑Lite). It is positioned as a faster, more affordable alternative to the current Gemini Pro models, though it offers slightly lower performance than the upcoming Gemini 3.5 Pro for tasks that require deep reasoning and high‑context understanding.
Performance Benchmarks
-
Coding & Agentic Tasks – Gemini 3.5 Flash outperforms Gemini 3.1 Pro on several benchmarks:
- Terminal‑Bench 2.1: 76.2 %
- MCP Atlas scaled tool use: 83.6 %
- Multimodal understanding (CharXiv Reasoning): 84.2 %
-
Speed – The model generates output tokens at roughly four times the rate of other frontier models, according to Google.
-
Long‑horizon Agentic Workflows – Google states that Flash can complete tasks that previously took weeks in “a fraction of the time,” reliably executing multi‑step workflows and coding tasks while maintaining frontier performance. Partners in banking and fintech have already used it to automate multi‑week processes.
Availability
- Default Model – Gemini 3.5 Flash is now the default model for the Gemini app and for AI Mode in Search worldwide.
- Personal AI Agent – The Gemini Spark personal AI agent, rolling out to testers, runs on Flash.
- Developer Access – The model is generally available through:
- Google Antigravity
- Gemini API in Google AI Studio and Android Studio
- Gemini Enterprise Agent Platform
- Gemini Enterprise
“It’s also now available to everyone in the Gemini app and AI Mode in Search,” Google said.
Safety Enhancements
Google has strengthened Gemini 3.5’s safeguards against cyber threats and CBRN (chemical, biological, radiological, nuclear) misuse, reducing the likelihood of generating harmful content or mistakenly refusing safe queries.
Future Plans
- Gemini 3.5 Pro – The higher‑performance Pro variant is already in internal use and is slated for public rollout next month. Pricing for paid tiers has not yet been disclosed.
References