Darkbloom on a Mac

Private AI inference marketplace on idle Apple Silicon Macs · AI / LLMs

Darkbloom (Eigen Labs) routes encrypted inference requests to idle Apple Silicon Macs; prompts stay unreadable to the Macs' operators; at roughly half typical API prices (pay-per-token, no subscription; public alpha). It has two sides and FrontierStack supports both. Calling it: the API is OpenAI-compatible at https://api.darkbloom.dev/v1 with a bearer key from console.darkbloom.dev; add a Darkbloom key in Model Costs and it appears in the assistant's model picker (gemma-4-26b, gpt-oss-20b; MoE models up to 239B, 128K context). Earning from it: an Apple Silicon Mac (macOS 14+, 48 GB+ RAM recommended) can serve models when idle; install the provider CLI with curl -fsSL https://api.darkbloom.dev/install.sh | bash (installs to ~/.darkbloom with a /usr/local/bin symlink; no launchd service), then darkbloom login, darkbloom models download and darkbloom start; darkbloom doctor verifies the machine, and earnings live at console.darkbloom.dev/earn. This pane detects the installed CLI and a running provider process. A live monitor watches Darkbloom's status page for coordinator incidents. Cloud caveats: even encrypted, prompts leave this Mac, so the prompt firewall and redaction apply as with any external provider; and the provider credentials in ~/.darkbloom (wallet_key, auth_token) are secrets; keep that folder out of backups you share.

Run Darkbloom with FrontierStack

Install or connect Darkbloom; then check its status, ports and certificate from FrontierStack. The same screen links to firewall checks, Malware Audit and backups where they apply.

Related in AI / LLMs

• Vane (Perplexica)
• MLX
• TurboFieldfare
• Cerebras
• MLX (Apple)
• vLLM
• llama.cpp Server
• Mojo (Modular MAX)