what u think ? #performance #asknostr
another ai assesment
Prompt Processing (2.75ms/token): Great.
Generation (75 t/s): Very Good.
Sustainability: High. This number is consistent for long runs.
Comparison: It is the expected "top-tier" performance for a 20GB+ VRAM GPU running local LLMs.
another ai assesment
Prompt Processing (2.75ms/token): Great.
Generation (75 t/s): Very Good.
Sustainability: High. This number is consistent for long runs.
Comparison: It is the expected "top-tier" performance for a 20GB+ VRAM GPU running local LLMs.
โ Pros & Cons
โ
Lightweight, fast, censorship-resistant, easy to self-host relays
โ
Works well with Bitcoin/Lightning, open standard, developer-friendly
โ No built-in spam filtering or ranking (relies on clients)
โ Evolving spec, no central entity to coordinate or enforce rules
โ Relay reliability varies; some charge for storage or bandwidth
NOSTR is essentially a protocol for open social data, not a company or app. If you want to dive into how to generate keys, run a relay, or pick a client, just let me know!