#openledger $OPEN A lot of AI models aren’t just not being made, but they’re not being paid for on a continuous basis.
The prices for general models are still being pushed down, while small teams often have fixed subscription fees, and the more custom versions they create, the higher the deployment and maintenance costs. The technology might work, but the numbers don’t always add up.
The cases of OpenLoRA and Aethir show that dynamically loading LoRA layers can bring up to 99% cost savings and run over 1,000 models on a single GPU. What’s really changing isn’t whether it can be deployed, but whether vertical models have the chance to shift from a one-time delivery to a long-term service.
If model access, API calls, and inference settlements are happening repeatedly, $OPEN might finally enter a real usage chain.
But the prerequisite still stands: after the price drops, will anyone really be willing to foot the bill?
@OpenLedger
The prices for general models are still being pushed down, while small teams often have fixed subscription fees, and the more custom versions they create, the higher the deployment and maintenance costs. The technology might work, but the numbers don’t always add up.
The cases of OpenLoRA and Aethir show that dynamically loading LoRA layers can bring up to 99% cost savings and run over 1,000 models on a single GPU. What’s really changing isn’t whether it can be deployed, but whether vertical models have the chance to shift from a one-time delivery to a long-term service.
If model access, API calls, and inference settlements are happening repeatedly, $OPEN might finally enter a real usage chain.
But the prerequisite still stands: after the price drops, will anyone really be willing to foot the bill?
@OpenLedger