giving ppq, with a (very buggy vibed tool) called routerly a try. Using tev:4b for routing decisions, a couple self hosted qwen-3.8:27b-q_8 slots, and ppq for access to big models, im hoping to drop my spend down to well under $2.00 per work-day at floor API prices.
ppq, only because it has a model pricing api endpoint, and no-kyc wallet. Eventually ill play around with TEE, but the work it does is all public/open source.
I found ollama cloud to be the most reasonable earlier this year, but their inference prices soared this summer so it became less affordable, and the GLM models started to suck after a while. I couldn't justify the $100/mo plan.
ChipTuner
npub1qd…3fqm7
2026-10-15 16:36:38 UTC
in reply to nevent1q…9cx9
Author Public Key
npub1qdjn8j4gwgmkj3k5un775nq6q3q7mguv5tvajstmkdsqdja2havq03fqm7Seen on
wss://nos.lolPublished at
2026-10-15 16:36:38 UTCEvent JSON
{
"id": "0f226af8e14a9d603fc229d5c8c63540a85cb81b1f7e68b0561303ab7526586b",
"pubkey": "036533caa872376946d4e4fdea4c1a0441eda38ca2d9d9417bb36006cbaabf58",
"created_at": 1792082198,
"kind": 1,
"tags": [
[
"e",
"1f60f9758b18d72995a3c66cb42d8d1b2943b78378326b55508c4dc578ee656d",
"wss://relay.nortis.tech/",
"root",
"e83b66a8ed2d37c07d1abea6e1b000a15549c69508fa4c5875556d52b0526c2b"
],
[
"e",
"d6228192a5199ccc5a1dba2ccf6331dd68d5dbbf277c82e9ec2a079f2dc0f868",
"wss://nos.lol/",
"reply",
"e83b66a8ed2d37c07d1abea6e1b000a15549c69508fa4c5875556d52b0526c2b"
],
[
"p",
"cdecc31c6d9406e9a7d6b0067412aa661d9d31c8035c3fd65c06301d1cac3b92"
],
[
"p",
"e83b66a8ed2d37c07d1abea6e1b000a15549c69508fa4c5875556d52b0526c2b"
]
],
"content": "giving ppq, with a (very buggy vibed tool) called routerly a try. Using tev:4b for routing decisions, a couple self hosted qwen-3.8:27b-q_8 slots, and ppq for access to big models, im hoping to drop my spend down to well under $2.00 per work-day at floor API prices. \n\nppq, only because it has a model pricing api endpoint, and no-kyc wallet. Eventually ill play around with TEE, but the work it does is all public/open source. \n\nI found ollama cloud to be the most reasonable earlier this year, but their inference prices soared this summer so it became less affordable, and the GLM models started to suck after a while. I couldn't justify the $100/mo plan. ",
"sig": "452b026c30a35d7ad20ae390f8a0faf98fddc23971c02a482a195a34ffe9ea074d81cc578f017d561680f97d49b7e6bca56eb3d637bf9401e46529ba4debef75"
}