Blackbox Enterprise Inference
Single-tenant hosting of an open-weight model the customer picks, or its own fine-tuned weights, on capacity reserved for that customer behind an OpenAI-compatible API, with in-memory processing, SSO, SCIM, role-based access and audit logs streamed to the customer's SIEM.
Find alternatives to Blackbox Enterprise Inference
What it does
- Model inference
- Model hosting
Features 6
- Dedicated single-tenant deploymentRuns your chosen open-weight model on capacity isolated from every other customer.
- Dedicated capacityReserved capacity sized to your traffic so throughput is unaffected by other customers.
- Zero data retentionPrompts, code and completions are processed in memory and discarded after the response.
- Bring your own weightsUpload fine-tuned models to a private registry and call them through the same endpoint.
- OpenAI-compatible APIExisting OpenAI SDKs work by pointing at your endpoint with the same request shapes.
- Identity and auditSAML SSO, SCIM provisioning, RBAC and tamper-evident audit logs streamed to a SIEM.
Sources
- Pricing
-
Enterprise "Custom", "Annual · PO burn-down", metered per token; the inference page cites a blended $0.44 per 1M tokens for NVIDIA Nemotron 3 Ultra.
- Description
-
Official documentation · 10 Oct 2026
"You choose the open-weight model, and we run it as a deployment dedicated to you and isolated from every other customer"; customers can upload fine-tuned weights to a private registry; SAML SSO, SCIM, fine-grained RBAC and tamper-evident audit logs streamed to a SIEM.
- What it does
- Constraints
-
Official documentation · 12 Oct 2026
Derived from the security facts on this page, from the same source.
- Security
-
Official documentation · 12 Oct 2026
Page: 'Training opt-out is on by default, and we enforce it with the upstream providers that we route to, wherever the provider API supports it'. / Page: 'Zero Data Retention keeps prompts, code context, and completions in memory only'.
- Sold within
-
"Every surface uses the same balance: Enterprise Inference, the Blackbox Router, the API, the CLI, and the Agents API." The page routes buyers to sales for "a deployment plan and per-token commit".
- Name
-
Official documentation · 10 Oct 2026
Enterprise Inference: "Your model. Your deployment. Yours alone."
- URL
- Deployment
-
Official documentation · 10 Oct 2026
Blackbox runs the deployment on its own infrastructure on reserved capacity dedicated to the customer; regions are not listed.
Also from Blackbox AI
-
Blackbox AI Infrastructure service
Model-inference gateway (Blackbox Router: 300+ open and closed models through one endpoint, zero data retention), Enterprise Inference (dedicated single-tenant GPU capacity) and an Agents API/CLI that sends coding agents to any repository over HTTP.
-
Blackbox Router API service
A gateway that puts 300+ open and closed models behind one OpenAI-compatible endpoint, routes each request to the best-performing provider with latency load-balancing and failover, and enforces zero data retention and no training on routed traffic.
-
Blackbox Agents API API service
An HTTP API that dispatches coding tasks to agents in isolated cloud sandboxes, streams every terminal event, can send one task to Blackbox, Claude Code, Codex and Gemini at once with a judging model picking the best result, and turns a successful run into a pull request.
-
Blackbox CLI AI agent
Blackbox's open-source terminal coding agent with plan mode, a multi-agent mode that runs agents in parallel, headless mode for CI/CD, and support for AGENTS.md, MCP servers, hooks and skills, metered per token against a Blackbox balance.
Open Blackbox AI in the directory
Something wrong here? Send a correction — quote this product id: blackbox-ai-enterprise-inference.