Arthur Hayes|11月 18, 2026 08:20
Many retail and enterprise FLOP miners to be ask us:
- What's the best hardware?
- How do I setup my hardware to optimise performance?
We cannot answer those questions. We can tell you the block reward, and that it's weighted by amount of inference processed measured by floating point operations (compute) meaning if in one block your miner processed 100 gigaflops out of a total of 200 gigaflops, you earn half of the miner block reward.
If you buy hardware that can serve a certain AI model, and that model is not in demand, you will process zero inference requests and earn nothing.
If you want to better forecast the types of models most in demand for agentic work, look at the most popular models on Open Router for example.
Worst case, try out your configuration, and adjust your setup based on actual results.
Share To
Timeline
HotFlash
APP
X
Telegram
CopyLink