Blackbox AI

Blackbox AI (BLACKBOX AI Technologies Inc.) is a high-trust AI coding and inference platform headquartered at 535 Mission Street, San Francisco. It serves developers and enterprises through four primary surfaces: a VS Code extension (4.2M+ installs) with inline code completion, AI chat, and multi-file agentic coding; a purpose-built AI-native IDE; a CLI for terminal-first coding agents; and an enterprise Blackbox Router that aggregates 300+ AI models (Claude, GPT, Gemini, Grok, Llama, Mistral, Nemotron, Kimi, Qwen, and more) behind a single endpoint with end-to-end encryption.

Blackbox's technical differentiation is in inference performance and trust architecture. It operates the #1 benchmarked provider for Nemotron 3 Ultra at 454 tokens/second (verified by Artificial Analysis, July 2026), 2.7x cheaper than the #2 fastest provider. On Enterprise, Blackbox runs dedicated single-tenant deployments isolated from all other customers with reserved capacity, PII anonymization before closed model APIs, and zero data retention contractually guaranteed in the DPA. A forward-deployed engineer is included at no cost to handle implementation, migration, eval harnesses, and routing optimization. The platform has a Microsoft Azure partnership for enterprise deployment in Microsoft-ecosystem environments.

For individual developers, Blackbox offers a genuinely free tier with unlimited chat and VS Code extension access. The Pro plan ($10/month) unlocks all 300+ models with MCP support. Enterprise is priced per token via annual purchase order commitments, with no per-seat fees — the token balance burns down as teams consume it across every surface (API, VS Code, CLI, Agents API) from one shared contract. Blackbox positions itself as the inference layer for teams that have outgrown the complexity and opaque pricing of managing multiple provider contracts separately.

Blackbox AI has a Freemium consumer tier and an enterprise-focused inference platform with distinct pricing structures.

The Free plan offers unlimited chat, VS Code extension access, and select model access with no credit card required. The Pro plan at $10/month provides access to all 300+ models with MCP support, full VS Code agent capabilities, and priority inference. Both are available at product.blackbox.ai/pricing.

Blackbox Enterprise Inference is priced per token, not per seat, via an annual purchase order commitment. The base rates (per 1M tokens, list rates): Nemotron 3 Ultra ($0.32 input / $0.80 output), Qwen3.7-Plus ($0.32 / $1.28), GLM-5.2 ($1.40 / $4.40), Kimi K2.7 Code ($0.74 / $3.50), MiniMax M3 ($0.60 / $2.40). Open-weight models (running on Blackbox-operated infrastructure) receive a 10%+ discount; closed models receive 5%+ off provider rates. Enterprise includes dedicated single-tenant deployment, data residency, SAML SSO, SCIM, RBAC, audit logs, zero data retention (contractual + DPA), PII removal before closed models, a forward-deployed engineer (included at no cost), Agents API, custom SLAs, and guaranteed throughput. Unused commitment expires at the billing period end with alerts at 75% and 90% burn-down.