Live drop
3.8 Flash
Google DeepMindDropped Sep 2, 202626 builds21 makers9 prompts
Long-horizon software engineering and agentic workflows with tunable thinking levels on 1M context.
Live drop
Google DeepMindDropped Sep 2, 202626 builds21 makers9 prompts
Long-horizon software engineering and agentic workflows with tunable thinking levels on 1M context.

About this release
Gemini 3.8 Flash is Google DeepMind's latest Flash-series LLM, improving on 3.7 Flash for complex coding, multi-step agentic tasks, and specialized reasoning. It supports text/image/audio/video inputs with text output, a 1M token context window, 64k output tokens, and low/medium/high thinking effort levels. Released as generally available on September 2, 2026.
Read the official Google DeepMind launch pageWorth opening first
Same brief, other models
All challenges →Challenge
Same brief across models — newer takes on the left, older on the right.
Challenge
Same brief across models — newer takes on the left, older on the right.
𝕏AppsMiMo-V2.6-ProLive reinforcement learning training dashboard for Xiaomi's MiMo 2.6 modelsby Fuli Luo
𝕏AppsRob's Design BenchA gallery comparing websites generated by AI models from a shared prompt.by Robin Ebers
𝕏AppsDGrid.AIDecentralized AI inference network verified on-chainby DGrid AIPaste the URL and Dropday.ai will credit you, track the build as it evolves, and put it on this drop.
Submit a build