Vercel added InclusionAI’s Ling 3.1 Flash 1019 to AI Gateway on October 30, 2026. Developers can reach it through the AI SDK, an OpenAI-compatible endpoint, and coding-agent integrations. The exact model identifier is inclusionai/ling-3.1-flash-1019.
The launch announcement positions the model for coding, multi-step reasoning, and agents that use tools. Its listed 1 million-token context window can accommodate long documents, code, and extended task histories.
The immediate catch is the pricing deadline. Vercel says requests begin charging on November 3, but the announcement and model listing supplied for this report do not state the post-promotion input and output rates.
What Is Free, and When Billing Starts
Vercel says requests are free until November 3, 2026. Those requests still appear in the spend dashboard with a full request trace, even though their inference cost is zero.
The announcement also explicitly says that charging begins on November 3. It gives no cutoff time or time zone, so developers should not assume they have the whole of November 3 in their local time zone.
| Billing detail | What Vercel documents |
|---|---|
| Promotional access | Free requests until November 3, 2026 |
| Start of paid usage | November 3, 2026 |
| Exact transition time | Not specified |
| Post-promotion token rates | Not specified in the opened launch materials |
| Request visibility | Free requests appear in the spend dashboard with full traces |
| Gateway inference fees | Vercel says it reflects provider pricing without markup or a platform inference fee |
These terms come from Vercel’s changelog. Its no-markup policy explains how it passes through pricing; it does not tell developers what this particular model will cost after the promotion.
I’d put a stop date on any unattended evaluation before November 3 unless the rate and cutoff have been confirmed. An agent loop left running can continue making requests after a promotional period ends.
Existing Prices Provide Context, Not a Forecast
Vercel’s InclusionAI catalog lists the new model as free for both input and output. It also lists paid rates for earlier models:
| Model | Listed context | Input per million tokens | Output per million tokens |
|---|---|---|---|
| Ling 3.1 Flash 1019 | 1M tokens | Free during promotion | Free during promotion |
| Ling 3.1 Flash | 262K tokens | $0.30 | $0.90 |
| Ling 3.0 Flash | 256K tokens | $0.02 | $0.06 |
The catalog labels Ling 3.0 Flash’s rates as 65% off, so those figures carry their own promotional condition.
For a concrete comparison, an evaluation consuming an aggregate 1 million input tokens and 100,000 output tokens across multiple calls would cost $0.39 at the listed, uncached Ling 3.1 Flash rates: $0.30 for input plus $0.09 for output. That is arithmetic using the older model’s published prices, not an estimate for 1019.
The useful conclusion is that developers cannot yet assign a dollar value to the new model’s free window. They can use it to establish task quality and usage patterns, then apply the actual paid rates when those are available.
Three Ways to Call Ling 3.1 Flash 1019
Vercel’s announcement documents three direct access paths. All use the same model identifier, but their authentication and client setup differ.
Use the AI SDK in a TypeScript App
For an application already using Vercel’s AI SDK with Gateway access configured, the model selection is a single string:
import { streamText } from 'ai';
const result = streamText({
model: 'inclusionai/ling-3.1-flash-1019',
prompt: 'List three reliability checks for a coding agent.',
});
for await (const text of result.textStream) {
process.stdout.write(text);
}
This example requests a streamed response and writes arriving text to standard output. It assumes the ai package and Gateway authentication are already configured; it is not a complete application setup.
The important configuration detail is the full identifier, including inclusionai/ and the 1019 suffix. Selecting the older inclusionai/ling-3.1-flash name would select a different catalog entry.
Use the OpenAI-Compatible Endpoint
Developers with an existing OpenAI Python client can configure it to send requests to Vercel’s Gateway endpoint:
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["AI_GATEWAY_API_KEY"],
base_url="https://ai-gateway.vercel.sh/v1",
)
stream = client.chat.completions.create(
model="inclusionai/ling-3.1-flash-1019",
messages=[
{
"role": "user",
"content": "List three reliability checks for a coding agent.",
}
],
stream=True,
)
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="", flush=True)
This requires the openai Python package and an AI Gateway key supplied through the AI_GATEWAY_API_KEY environment variable. The key belongs to Vercel’s Gateway, even though the client library is OpenAI’s.
The base URL determines where the request goes; the model string selects the catalog entry. This is a convenient route for applications already structured around Chat Completions, though the sample alone does not verify compatibility with every optional API feature.
Use fx From a Project Directory
Vercel also documents an fx workflow for macOS and Linux on x86_64 and arm64 systems. Its installation and launch commands are:
curl -fsSL https://fx.sh/setup.sh | bash
fx
The installation command downloads and executes a shell script. I’d inspect that script before running it, particularly on a machine containing production credentials.
Run fx from the project directory. With no saved credentials, it opens a connection picker offering Sign in with Vercel or Add an API key. Inside fx, these commands reopen setup and select the model:
/setup
/model inclusionai/ling-3.1-flash-1019
Vercel says signing in exposes the full AI Gateway catalog in the model picker. All three access paths above are documented in the launch announcement.
Use It in Claude Code, Codex, or Cursor
Vercel names Claude Code, Codex, and Cursor among the coding agents supported through its Gateway setup workflow.
The documented entry point is:
npx vercel ai-gateway setup
Frequently Asked Questions
3 questions
1When Does Ling 3.1 Flash 1019 Start Charging?
Vercel says charging begins on November 3, 2026. The announcement does not specify a time zone or exact cutoff time.
