Darkbloom provides a privacy-focused AI inference network by routing encrypted requests to hardware-verified Apple Silicon machines. It is designed for developers seeking cost-efficient, private AI inference, and for Mac owners interested in earning revenue from their idle Apple Silicon hardware. The platform emphasizes keeping user prompts and inference data hidden from both the coordinator and the hardware operator, addressing privacy concerns when running inference on machines not directly controlled by the user.
The service achieves privacy using several independently verifiable layers. Requests are encrypted before leaving the developer's application, and only the matched provider's hardware-bound key can decrypt them. Each provider uses a key generated within Apple's tamper-resistant secure hardware, with an attestation chain tracing back to Apple's root certificate authority. The inference process is secured at the OS level, blocking debugger attachment and memory inspection, while responses are signed by the specific machine that produced them, allowing users to verify the hardware origin.
Developers can access Darkbloom's inference capabilities without adopting a new SDK, as the platform is compatible with existing OpenAI clients; only the base URL needs to be changed. The API supports streaming in the OpenAI format and offers access to selected large mixture-of-experts models, with examples including models up to 239 billion parameters. For Mac owners, installing the provider involves running a terminal command to download and configure the background service, which requires no dependencies and auto-updates. Operators can choose when their Macs are available for inference jobs and, during the public alpha, keep 100% of the inference revenue generated.
Pricing is structured on a pay-per-token basis with no subscription or minimums, and model prices are set around 50% lower than typical API-provider rates for comparable models. The platform leverages the idle capacity of Apple Silicon devices, offering a network that combines privacy guarantees with reduced costs for developers and earning opportunities for hardware owners.
Darkbloom sits in PulseGate's Inference & model serving category. Enabling developers to run private, cost-effective AI inference without exposing sensitive data to third-party operators. It is built as a B2B product for developers. It runs on the web, the command line, and API.
Darkbloom first shipped in 2026. The project is developed in the open on GitHub with 220 stars and 758 commits in the last 90 days. Key capabilities include encrypted inference, Apple Silicon support, and operator-blind privacy. It exposes integrations via a public API.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match