Ling 3.0 Flash Sante is now available on AI Gateway for free install

What shipped

Vercel AI Gateway has added Ling 3.0 Flash Sante from inclusionAI to its model roster, and it is free to use through October 4. If you are building health or medicine applications and have been looking for a capable model without an upfront cost, this is a reasonable moment to evaluate what the model can do before the free window closes.

The model is a Mixture-of-Experts architecture with 124 billion total parameters and roughly 5.1 billion active per token. It supports a 256K token context window and includes function calling. That combination makes it suitable for long clinical documents, multi-turn patient-facing conversations, and tool-augmented workflows where the model needs to call external APIs mid-conversation.

Ling 3.0 Flash Sante is a health and medicine-focused variant of Ling 3.0 Flash, produced by inclusionAI. It is now live on Vercel AI Gateway. The model is free to use through October 4. After that date, standard pricing applies, though the exact post-free-tier cost has not been specified in the announcement. The model is accessible through the AI Gateway interface without any separate arrangement with inclusionAI directly.

What improved

The main practical change is that AI Gateway now routes to this model alongside whatever other models you already use there. You do not need a new account, a separate API key from inclusionAI, or any special approval to start calling it. Because it sits behind the same AI Gateway abstraction, your existing logging, rate limiting, caching, and observability tooling applies automatically. The 256K context window is notably larger than many general-purpose models, which matters when you are passing in long medical histories or regulatory documents. Function calling support means you can wire the model into retrieval pipelines, scheduling systems, or any external service you expose as a tool without post-processing the output yourself.

Start by opening your Vercel dashboard and navigating to the AI Gateway section. If you do not already have a gateway configured, create one. Select the model provider list and locate Ling 3.0 Flash Sante from inclusionAI. Add it to your gateway configuration and save.

In your application code, update the model identifier string you pass to the AI Gateway client to point at Ling 3.0 Flash Sante. The exact identifier string is displayed in the gateway configuration panel after you add the model. No SDK version changes are required on the client side as long as you are already using the AI Gateway SDK or the standard fetch-based approach with the gateway endpoint.

What you gain

If you want to test the model before touching production code, create a separate gateway environment in Vercel, point it at Ling 3.0 Flash Sante, and run your existing integration tests against that environment first. Once you are satisfied with the responses, update your production environment variable for the model identifier.

For teams using infrastructure as code to manage Vercel resources, update the relevant configuration file to include the new model identifier, run a preview deployment, validate outputs, and then promote to production. Because the model runs behind the shared gateway endpoint, there is no infrastructure provisioning step on your end.

How to get it

The free access window closes on October 4, so any workload you move to this model before then will need a plan for what happens after. Check whether your current AI Gateway plan covers the cost of continuing to use the model, and set a calendar reminder a few days before the cutoff so you have time to either confirm billing arrangements or route traffic elsewhere before you are surprised by charges or a broken integration.

The 256K context window is an opportunity but also a cost consideration once billing resumes. Very long prompts cost more per call. If you are currently truncating documents to fit smaller models, you may expand context significantly and then see a larger bill than expected after October 4. Profile your average prompt length during the free period so you have a realistic baseline.

Function calling behavior can vary between models even when the interface is standardized. Run your existing function-calling tests explicitly against Ling 3.0 Flash Sante rather than assuming it will behave identically to whatever model you replaced. Pay particular attention to how it handles ambiguous tool selections and whether it correctly declines to call a tool when none is appropriate.

What to watch next

Because this model is health and medicine focused, responses to general-domain prompts may differ in tone or content from a general-purpose model. Test representative prompts from your actual use case rather than generic benchmarks.

Keep an eye on the Vercel AI Gateway changelog and the inclusionAI announcements for pricing details after October 4. The free period is a useful evaluation window, but production readiness requires knowing the ongoing cost. Watch also for any updates to the model itself, since a Flash Sante designation suggests this is a version within a larger model family and further variants or updates may follow.