NEAR AI Cloud introduced full compatibility with OpenAI APIs on December 3, 2025. Developers can migrate existing applications to the NEAR infrastructure by updating a single URL and using a NEAR AI API key. The system supports Python and JavaScript SDKs, chat completions, legacy completions, tool calling, and structured outputs.
The platform utilizes Trusted Execution Environments (TEEs) to ensure data privacy. Models, such as GLM 5.2 variants, operate within Intel TDX and NVIDIA confidential computing hardware. Each request includes cryptographic attestation to verify that the computation occurred within a secure enclave without external interference.
In July 2026, NEAR implemented a staking-based payment model for AI compute credits, allowing users to stake NEAR tokens for capacity. Between July and September 2026, the company expanded its model catalog and added support for integrations with LiteLLM, Aider, and Continue.
While the infrastructure provides privacy, it introduces specific trade-offs. TEE-based inference may result in higher latency compared to standard GPU compute, and the staking payment mechanism exposes enterprise users to potential volatility in NEAR token prices.
