Ling-3.0-tiny lands free on Vercel's AI Gateway through August 14

InclusionAI released Ling-3.0-tiny on August 6, and Vercel dropped it straight into the free slot on its AI Gateway, per the Vercel changelog. For builders testing agent loops on a budget, that means a tool-calling model with 256K of context is free to try through August 14.

Ling-3.0-tiny runs on 1.3B active parameters out of 7.9B total, a size built for constrained hardware and fully local deployment. It switches between a Thinking mode for reasoning and an Instant mode for fast replies, and it keeps native function calling and prompt caching even at that small active-parameter count. Vercel demoed it running obsidian-cli against a local notes repo, an agent task that usually needs a bigger model to hold context and call functions reliably.

That's the part worth noticing. Most models this small give up tool calling or context length to hit a low parameter count. Ling-3.0-tiny keeps both, so it can run an actual agent loop on modest hardware instead of only answering chat prompts.

Vercel is giving it away through 8:00am PT on August 14; the changelog doesn't say what fills the slot after that. The real test isn't the benchmark numbers, it's whether builders actually run a 1.3B-active-parameter agent loop while it costs nothing, or wait to see if it holds up once the meter's back on.