Projects / Darkbloom

About

Darkbloom

Darkbloom

API Offchain

An off-chain private-inference network built around Apple Silicon. This dashboard reports source-labelled application traffic, capacity, model pricing, and aggregate hardware trust signals. Its B200 Index uses only public attestation rows that identify both an Apple chip and its GPU-core configuration.

darkbloom.dev/ Docs GitHub

Overview

B200 Index ? 20.67 ▼ 6.8% 7d
Throughput ? 2.41Btok/day ▼ 9.5% 7d
Community Score ? 19.50/ 100

Apple Silicon Capacity & Trust

Latest public aggregate snapshot; not GPU inventory or uptime
Active Providers1,162
Reported Capacity (tok/s)19,976.26
Network Utilization6.99%
Active / Queued Requests385 / 0
Model Capacity (tok/s)86,307.26
Cumulative Requests114,666,838
Cumulative Tokens400,871,447,414
24h Requests / Tokens2,704,151 / 0
Apple GPU Cores40,137
CPU Cores18,326
Unified RAM (GB)109,214.00
Memory Bandwidth (GB/s)527,328.00
Power Under Load (W)176,020.00
Hardware-attested Providers1,168
B200 Index (Attested Macs)20.671638
Theoretical Throughput (tok/day)2.41B
Secure Enclave Providers1,168
Code-attested Providers1,028

Darkbloom's aggregate Apple GPU-core total is not a B200 Index. Separately, public attestation rows that identify chip, GPU-core count, and per-Mac unified RAM are aggregated with the documented Apple Silicon coefficients. The theoretical throughput applies the same 70 GB reference workload used for other networks, so only Macs with at least 96 GB RAM contribute; raw B200 Index remains independent of this gate. This is attested-machine capacity only—not uptime, delivered capacity, or a conversion of RAM/aggregate core totals. Darkbloom source ↗

Inference

Socials