rotative
lesswrong ai · May 22, 2026

Google releases Gemini 3.5 Flash for agentic workflows

google released gemini 3.5 flash, a hybrid model optimized for speed and agentic tasks, outperforming 3.1 pro on benchmarks like terminal-bench while running 4x faster than frontier models.

read in 简体中文 · 繁體中文 · 日本語 · 한국어

google launched gemini 3.5 flash as a hybrid model designed specifically for high-frequency, iterative agent workflows, balancing frontier-level intelligence with execution speed. the model demonstrates strong performance on internal agentic and coding benchmarks like terminal-bench, where it surpasses 3.1 pro, and achieves speeds up to 12x faster when integrated with google's antigravity harness. despite being labeled as a flash model, the pricing is higher than previous versions, positioning it closer to expensive models like opus 4.7 while retaining the speed benefits necessary for sub-agent collaboration. independent reviews noted significant weaknesses, including poor performance on the you're absolutely right sycophancy benchmark and a january 2025 knowledge cutoff that limits current awareness. google confirmed that the more powerful 3.5 pro model is scheduled for release next month.

read at source ↗

← rotative