LG AI Research announced free hosted access to K-EXAONE 2.5 on November 9, 2026, giving developers and researchers a way to explore the model without running it on their own servers.
Access is through an API—a connection that lets software send requests to the model. LG’s same-day platform changelog confirms that the service is currently free.
LG’s platform terms restrict access to non-commercial use. They also prohibit using the model’s output commercially or letting third parties send requests through an integrated product or service. The announcement’s invitation to companies and service developers should not be read as permission to launch a customer-facing AI service.
What the hosted offer provides
LG’s model documentation lists Korean and English text support, chat and code generation. It also lists tool calling, which lets the model request actions from other software, and structured outputs, which let it return data in a specified format.
For a permitted non-commercial experiment, a personal learning project—such as asking for help with code you wrote—offers a concrete starting point. These are documented capabilities, not a guarantee that the answers will be correct for a particular task.
Tool support does not make the API a complete automation service. LG’s agent guide places execution of tool requests in separate software, which then returns the results to the model. Free model usage therefore does not eliminate the work or other costs of building an application.
This is a different access route from self-hosting K-EXAONE 2.0, covered previously by RohitAI. That earlier model’s hardware requirements and licensing should not be carried over to the hosted 2.5 offer.
Free does not mean unlimited
The rate-limit guide says excess requests receive a “Too Many Requests” error, or HTTP 429. Developers can read their limits and remaining allowance in the metadata returned with each response. The guide’s numerical example is illustrative, not a guaranteed quota for every account.
Nor is the current price a permanent commitment. LG’s terms allow future charging and changes to, or discontinuation of, a model. No end date for free access or future price schedule was identified in the reviewed pages.
The terms also set a minimum age of 19 and say API keys expire after 30 days. The reviewed documentation does not establish signup availability in every country.
Before submitting material, review the privacy policy and consent settings: LG describes use of prompts and responses for model training as optional, consent-based processing. Its service terms separately prohibit submitting another person’s personal information.
The practical distinction is between the price of access, permission to use it and available capacity. For a small business, that means clarifying the intended use with LG before treating this free offer as a replacement for a commercial AI service.