@ollama Therefore, the current usage is likely routed, and to prevent losses, they replicated the upstream pricing. Once the model can be served on their own cluster, things should return to normal.
@ollama I think Ollama's usual model is to provide inference services with their own cluster, and Kimi K3 has just released the weights, so they probably haven't had time to convert them to offer their own service.
@kimmonismus He now still preaches to us about the dangers of open source, but once the laws he advocates pass through Congress, he will stop saying such things. By then, he will become just like the patent troll Dolby, directly suing anyone who threatens his business everywhere.
@OpenRouter The AA Index for the Nemotron 3 Ultra should be 38, and given its performance, it basically doesn't belong on this list. You should wait until they release the Nemotron 4 Ultra before recommending it to people, the current version isn't really suitable for serious tasks.