Google launches Gemini 3.6 Flash with better agent benchmarks and lower token use
DevToD4 · x · 2026-07-22
Google DeepMind says it is rolling out three new models for agents:
- Gemini 3.6 Flash: higher quality than 3.5 Flash at the same cost, with fewer tokens.
- Gemini 3.5 Flash-Lite: a low-cost option for document processing and agentic search.
- Gemini 3.5 Flash Cyber: a cybersecurity model for finding and patching critical vulnerabilities.
The promo chart highlights gains for Gemini 3.6 Flash over 3.5 Flash, including:
- DeepSWE: 37% → 49%
- ML engineering: 42.6% → 63.9%
- Knowledge work: 965 → 1421
- Computer use: 76.2% → 83.0%
Related event: Google Launches Gemini 3.6 Flash to Boost AI Agents(2 posts)→
More from Models
- OpenAI Presence appears to be a limited enterprise deployment product — btibor91 · 2026-07-22
- Kimi K3 is called roughly equivalent to Opus 4.8 on ALE-Bench — scaling01 · 2026-07-22
- Tencent Hunyuan’s Hy3 ranks #25 in Agent Arena and #2 among open models in Frontend Code Arena — arena · 2026-07-22
- Frontier models are becoming planners, while cheap models handle execution — bigdata · 2026-07-22
- Three prompts behind 47 n8n and Claude agents — Aiden_Tech_Ai · 2026-07-22
- Reddit user says Microsoft’s Mageflow 4B blocks famous-character prompts — krigeta1 · 2026-07-22