NEAR AI Cloud has added Moonshot AI’s Kimi K3.1 Pro to its model catalog, giving developers another frontier-scale reasoning model through the platform’s unified inference interface. The official Kimi K3.1 Pro listing sets pricing at $2 per million input tokens, $10 per million output tokens and $0.20 per million cached-read tokens. The model supports a 1 million-token context window and accepts multimodal workloads, including native image and video understanding.
Moonshot AI launched Kimi K3.1 Pro on October 23 as the first model in its new K3.1 family and described it as its most capable 3T-class open-weight model to date. Moonshot positions the model for long-horizon coding, 3D creation, financial research and agentic knowledge work, while NEAR AI now provides another access route through an OpenAI-compatible API. Those workload descriptions are vendor claims rather than independent measures of performance.
Kimi Uses NEAR AI’s Incognito Route
The execution model is an important part of the integration. Kimi K3.1 Pro is classified under NEAR AI’s Incognito tier rather than its hardware-confidential model tier. Inference runs on a partner provider’s infrastructure, while NEAR AI routes requests using shared provider credentials so the upstream service does not receive the user’s NEAR AI account credentials. The prompt itself is still processed by that provider, and NEAR AI explicitly states that model-level TEE attestation and client-to-model end-to-end encryption are unavailable for Kimi K3.1 Pro.
That distinction mirrors the architecture behind NEAR AI’s OpenAI-compatible inference layer, where developers can access different execution models through a common API without receiving identical privacy guarantees. Confidential open-weight models can run inside attested hardware, while externally hosted models trade that execution guarantee for access to provider-operated frontier systems.
The same separation applies to other additions such as GPT-6.1 Sol inside NEAR AI Cloud. That model also uses an Incognito route, while supported open-weight alternatives can run through NEAR AI’s confidential infrastructure. The growing catalog therefore represents a multi-model access layer rather than a claim that every available model is privately hosted by NEAR AI.
Model Choice Expands Across a Common Interface
NEAR AI currently lists more than 60 models spanning text, image generation, embeddings, reranking and transcription. Kimi K3.1 Pro joins that catalog with one of its larger context windows, while developers can call it through the same /v1/chat/completions interface used by other supported models. A common API can reduce integration friction when applications need to switch between models based on cost, capability, context size or privacy requirements.
Distribution is also expanding beyond NEAR AI’s direct interface. The platform recently became an inference provider through OpenRouter, although that route introduces its own privacy boundary because requests pass through OpenRouter before reaching the selected provider. Model availability and confidential execution therefore remain separate properties even as NEAR AI increases the number of ways developers can reach its infrastructure.
Kimi K3.1 Pro broadens that choice with explicit commercial pricing and a million-token context window, but the meaningful product distinction lies in how developers balance model capability against execution provenance and privacy guarantees. Its addition expands NEAR AI’s model-selection layer without turning Moonshot’s externally operated inference into the hardware-attested confidential compute available elsewhere in the platform.