Open Source

Step 3.5 Flash is a beast?

A new open model is shocking users by rivaling top-tier closed-source AI.

Deep Dive

A viral user report claims Step 3.5 Flash is a "beast" for complex, agentic tasks, outperforming Gemini 3.0 Pro. The user tested it on a 90k context task, noting its speed and sanity, and was so impressed they double-checked they weren't using a frontier model like Claude 4.6. The model also performed well on OpenCode and Claude Code, sparking debate about open-weight model supremacy.

Why It Matters

This suggests open-source models are rapidly closing the performance gap with expensive, proprietary leaders.