Performances
RD Popularity Index
RD Growth Index
Comparison
Developer
Follow Updates
Follow Lists
People
About Darkbloom
Darkbloom is a decentralized private AI inference network built by Eigen Labs. It connects idle Apple Silicon Macs into a distributed inference cloud, letting Mac owners run a simple CLI tool to earn income from their spare computing power. Users access the network through an OpenAI-compatible API at roughly half the cost of centralized providers.
The core pain point it addresses is twofold: centralized AI inference is expensive and creates privacy risks, while millions of Macs sit mostly idle with powerful unified memory. Darkbloom turns this idle capacity into a cheaper, privacy-first alternative.
Its key competitive narrative is privacy through hardware-level security. Inference runs in-process via MLX with no subprocess or local server; debuggers are denied at the kernel level; requests are end-to-end encrypted; and a four-layer hardware attestation chain (Secure Enclave, MDM checks, Apple Managed Device Attestation, APNs code-identity) ensures node operators cannot read prompts or responses. The only plaintext exposure is transiently inside the coordinator's hardware-encrypted confidential VM memory, never logged or retained.
In the past six months (March to September 2026): the project launched on April 14; by late August it processed 5 billion tokens daily and 164 billion cumulatively; online Macs grew from zero to over 250; a single Mac earned $120–200 per month. It supports models like Qwen3.6-35B-A3B and Gemma 4 26B, with streaming, tool calling, vision input, and a structured 404 for unimplemented endpoints.
Important caveat: Darkbloom is under active development, has not been security-audited, and is not production-ready. It is explicitly labeled as for testing only. Running high duty cycles increases electricity costs and hardware wear, and income is an estimate, not guaranteed.
Darkbloom, a decentralized private AI inference network by Eigen Labs, entered public alpha in mid-April 2026, initially supporting gpt-oss-20b and Gemma 4 26B models on idle Apple Silicon Macs. On June 23, it launched on OpenRouter with free inference, quickly gaining traction. A key milestone came on July 27 with a major MLX continuous batching redesign, boosting prefill speed 3-4x, alongside the MLX.fast optimization contest. Network scale surged from 250 to 1,300 Macs, processing over 164 billion tokens cumulatively, with daily volume reaching 5 billion. On August 20, it transitioned from OpenRouter's free tier to paid status, generating approximately $102,000 in annualized revenue. Architecturally, Darkbloom employs encrypted routing, hardware trust roots via Apple's Secure Enclave, and hardened execution, ensuring privacy even from node operators. Its coordinator remains a trusted single point, explicitly disclosed, while the project avoids token emissions, relying purely on real inference demand. These milestones highlight its unique bet on consumer hardware for AI inference and its sustainable, privacy-centric design.
Darkbloom, EigenLayer's decentralized private inference network, made significant strides from Mar to Sep 2026. In mid-April, the research preview launched on Apple Silicon, with 9 Macs online providing 600+GB unified memory. It achieved verifiable private inference via Secure Enclave and hardware attestation, with providers keeping 100% of revenue (0% platform fee) and prices ~50% below OpenRouter. By July, the Swift migration was completed, focusing on large-scale MoE text inference while deprioritizing image/STT. In August, the network expanded to 250 devices, generating ~$45K monthly revenue, with top providers earning $120-200/month. Future roadmap includes: AVS-style node verification via restaking, demand-matching optimization to sustain provider utilization, potential tokenization aligned with EigenLayer's "build-first" approach, expanding to image/video modalities, and formal security audits. The team emphasizes privacy-first design fitting 2026's compliance-driven AI shift.
Darkbloom was created by Eigen Labs as a research initiative; it has no independent co-founders. Core leadership: Sreeram Kannan, Founder & CEO of Eigen Labs, former associate professor at the University of Washington, expert in information theory and blockchain scaling. Current key executives: Johnathan (JT) Rose, Chief Growth Officer at Eigen Labs, driving Darkbloom's growth and ecosystem partnerships. Core engineer: Gajesh Naik, who led Darkbloom's development and public launch. Pratik Gandhi, part of Eigen Labs' open frontier research and coordination network efforts, is involved in promoting Darkbloom. Darkbloom remains a research project without a formal executive structure.