Third-Party Models & AI Notice
The Services may route requests to multiple AI models. The model used may depend on user selection, plan, location, feature, availability, safety controls, latency, cost or system routing. Not every model is used for every request. Additionally, while you are permitted to select which model you intend to use, other models may also be used in order to ensure that the user has the most efficient route to achieving the task as thoroughly and completely as possible.
Model/provider schedule
| Model/feature | Provider/operator | Route | Terms/notes |
|---|---|---|---|
| GPT family (GPT-5.6 Sol, Terra, Luna) | OpenAI | Vercel AI Gateway | Commercial API terms apply. OpenAI also provides input safety moderation (direct API) and text embeddings. |
| Claude family (Claude Fable 5, Opus 4.8/5, Sonnet 5, Haiku 4.5) | Anthropic | Vercel AI Gateway (primary); Anthropic API (automatic fallback) | Commercial API terms apply; zero-data-retention routing available on supported Claude models. |
| Gemini family (Gemini 3.1 Pro, 3.6 Flash); Imagen 4 (images); Veo 3.1 (video) | Vercel AI Gateway | Gemini API terms apply. | |
| Grok family (Grok 4.3, 4.5; Grok Imagine image/video) | xAI | Direct xAI API (Gateway fallback) | xAI API terms apply; not a zero-data-retention route. |
| OpenServe Genius; OpenServe Nexus (Company-branded) | Openserve Holdings, LLC, using the third-party models described in the Company Model disclosure below | Third-party inference via Vercel AI Gateway and provider APIs (no self-hosting) | Company branding does not displace underlying license/attribution obligations. |
| Additional text models (DeepSeek, Meta Llama, Mistral, Alibaba Qwen, Z.ai GLM, Sakana, MiniMax, Perplexity Sonar, Amazon Nova, Cohere Command) | Respective providers | Vercel AI Gateway | Provider API terms apply as routed through the gateway; availability varies by plan. |
| Additional image/video models (Black Forest Labs FLUX, ByteDance Seedream/Seedance, Recraft, Kling) | Respective providers | Vercel AI Gateway | Availability varies by plan and data-retention settings. |
Important terms
- ·Outputs may be inaccurate, incomplete, biased, unsafe, insecure or non-unique; users remain responsible for review and use.
- ·Model providers/routes may change; substitution may occur when a selected model is unavailable or plan settings allow routing.
- ·Input may be transmitted to the provider/operator used for a request; data handling is described in the Privacy Notice and DPA.
- ·Provider flow-down restrictions apply only where contractually or substantively applicable; provider consumer terms should not be incorporated by assumption.
- ·Open-source/third-party model licenses may impose attribution, notice, redistribution, naming or acceptable-use requirements that survive Company branding.
- ·Automated routing, moderation, embeddings, reranking, search and other supporting AI components may differ from the user-facing model selection.
Company Model disclosure
Openserve may display one or more models under a Company brand (currently OpenServe Genius and OpenServe Nexus). Company-branded models are not separate foundation models: they consist of Company-configured routing, prompting, safety and orchestration over unmodified third-party models accessed by API. OpenServe Genius currently routes each request to any one of the engines listed above depending on the task. The Company does not fine-tune, retrain or host the underlying models; the Company’s customizations consist of model selection and routing logic, system prompts, safety filtering and guardrails, tool orchestration and usage controls. The engines behind Company-branded models may change as quality and availability evolve, and the current engine list is maintained on this page. The underlying models remain subject to their applicable licenses and notices.
Related
See What is Genius? for the product view of the Company-branded models, and our Privacy Notice for how prompts and responses are handled.