NEAR AI Cloud launched full OpenAI compatibility on December 3, 2025, allowing developers to move existing OpenAI client code to NEAR infrastructure by changing one URL and replacing the API key. The service supports chat completions and legacy completions, along with Python and JavaScript SDKs, tool calling and structured outputs. Requests can pass through a central gateway or go directly to model-specific endpoints running in Trusted Execution Environments (secure hardware areas for private computation). NEAR says its TEE-hosted models, including GLM 5.2 variants, use Intel TDX and NVIDIA confidential computing hardware, with each request accompanied by a complete cryptographic attestation. Around July 2026, NEAR added staking-based payments for AI compute credits, allowing developers and companies to stake NEAR tokens for access to computing capacity. The model catalog expanded between July and September 2026, with documented integrations for LiteLLM, Aider and Continue. The offering extends NEAR Protocol's developer-focused layer-1 blockchain, account abstraction and usability strategy into AI infrastructure. Its privacy model can add latency versus standard GPU computing, while token-based payments expose enterprise customers to NEAR price volatility.