REDDIT · REDDIT// 9d agoMODEL RELEASE

StepFun launches Step 3.5 Flash 2603

StepFun has rolled out Step 3.5 Flash 2603, an updated version of its Step 3.5 Flash model for Step Plan users. The release adds a low-think mode and targets coding and agent frameworks, with StepFun claiming similar reasoning quality in high mode while reducing token use by about 14%, and up to 56% in low-think mode. The update is clearly aimed at high-frequency, lower-complexity agent tasks where speed and cost matter more than exhaustive reasoning.

// ANALYSIS

This looks like a practical efficiency upgrade rather than a brand-new model reset.

–Low-think mode is the main feature: it gives users a cheaper, faster option for routine agent steps.
–The coding and agent tuning suggests StepFun is optimizing for real workflows, not just benchmark vanity.
–If the token-savings hold in practice, this becomes a strong fit for orchestration layers and multi-model stacks.
–The launch matters most to existing Step Plan users and OpenClaw/OpenRouter adopters who run lots of short, repetitive tasks.

// TAGS

stepfunllmmodel updateagentscodinglow-think-modeopen-source

DISCOVERED

9d ago

2026-04-02

PUBLISHED

9d ago

2026-04-02

RELEVANCE

8/ 10

AUTHOR

tarruda