c0mpute.com

What It Is

c0mpute is an open-source, CLI-first decentralized GPU compute marketplace maintained by Profullstack, Inc. Buyers can pay Workers with idle GPUs to run tasks, while Workers register their hardware and earn money by accepting jobs. It currently includes three built-in modules: transcode for FFmpeg video transcoding, coinpay for DID identity and escrow payments, and infernet for AI/LLM inference. The project is licensed under MIT and emphasizes having no centralized backend, peer-to-peer connections, and modular extensibility.

Core Capabilities

For AI workloads, infernet can run LLM and other ML inference tasks on Worker GPUs. Examples include qwen and llama-3.1-8b, with support for listing network models, benchmarking, and pinning runtime image hashes to improve reproducibility. On the transcoding side, it supports H.264, HEVC, AV1, HLS, and 4K transcoding, with hardware acceleration when available. For quality validation, transcoding uses ffprobe and optional VMAF; AI inference uses output schema checks and repeated sampling runs. The platform also provides DID-signed task manifests, Worker reputation metrics, a TUI dashboard, plugin installation, and c0mpute doctor self-checks.

Pricing and Trial

c0mpute has no fixed pricing. The CLI, DID creation, and network exploration are free; users only pay when submitting tasks. Buyers set a maximum price with --max-price, Workers quote their own prices, and if a quote is below the cap, the task is executed. After completion and verification, funds are released through CoinPay escrow. The project states that there are no platform fees and that Workers receive 100% of what buyers pay. Early testing examples suggest LLM inference costs around $0.01–$0.10 per run, 1080p HLS transcoding around $0.50–$1.25, and 4K AV1 around $2–$5, but these are not guaranteed prices.

Pros and Cons

The strengths are open-source transparency, a clear command-line workflow, no platform commission, and coverage of two GPU-intensive use cases: AI inference and video transcoding. Its DID, escrow payment, and verification mechanisms also make decentralized task collaboration more practical. The limitations are also clear: the project is still pre-mainnet, the live network is in early testing, and Worker supply, pricing, task latency, and stability remain uncertain. It is primarily CLI-based, making it less friendly for non-technical users, and the formal terms of service have not yet completed legal review.

Best Fit and Access from China

c0mpute is better suited to developers familiar with terminals, Docker/GPU environments, and open-source toolchains; users running batch AI inference; video processing teams; and individuals or organizations looking to monetize idle GPUs. The available materials do not mention a Chinese interface, Chinese documentation, network connectivity from China, or payment methods, so accessibility from China is unknown. Payments are only described as being handled through CoinPay escrow, with no indication of whether common Chinese payment tools are supported. If you need a mature managed experience, alternatives to compare include RunPod, Vast.ai, Replicate, Together AI, and Akash Network.