Just tried qwen3.6;27b with llama.cpp and a bunch of optimizations and yeah much much better. 128k context instead of 64k and significantly faster than stock ollama. It fixed in 5 minutes what it couldn't do in a day.
Still need to try Gemma though.
Daniel Wigton
npub1w4…q30zp
2026-05-20 01:23:25 UTC
in reply to nevent1q…gxl2
Author Public Key
npub1w4jkwspqn9svwnlrw0nfg0u2yx4cj6yfmp53ya4xp7r24k7gly4qaq30zpSeen on
wss://nos.lolPublished at
2026-05-20 01:23:25 UTCEvent JSON
{
"id": "1afd4dffea4c9f176250466610b9c11f7f075d4e38692c10d72e449d65347000",
"pubkey": "75656740209960c74fe373e6943f8a21ab896889d8691276a60f86aadbc8f92a",
"created_at": 1779240205,
"kind": 1,
"tags": [
[
"alt",
"A short note: Just tried qwen3.6;27b with llama.cpp and a bunch ..."
],
[
"e",
"694e18c193755567577504247b85ffe395587b123551ad6eea32b894346dee85",
"wss://theforest.nostr1.com/",
"root",
"036533caa872376946d4e4fdea4c1a0441eda38ca2d9d9417bb36006cbaabf58"
],
[
"e",
"29685356379de80a8cee593b455d984819245a25bfab8e503736db1f4db05465",
"wss://theforest.nostr1.com/",
"",
"75656740209960c74fe373e6943f8a21ab896889d8691276a60f86aadbc8f92a"
],
[
"e",
"1ee333e294cb8eb45fcb2ba53ca57874a0e7fc7762fb0508efa492914df1a7a7",
"wss://theforest.nostr1.com/",
"reply",
"036533caa872376946d4e4fdea4c1a0441eda38ca2d9d9417bb36006cbaabf58"
],
[
"p",
"036533caa872376946d4e4fdea4c1a0441eda38ca2d9d9417bb36006cbaabf58",
"wss://relay.primal.net/"
],
[
"p",
"75656740209960c74fe373e6943f8a21ab896889d8691276a60f86aadbc8f92a",
"wss://theforest.nostr1.com/"
],
[
"r",
"https://llama.cpp/"
],
[
"client",
"Amethyst"
]
],
"content": "Just tried qwen3.6;27b with llama.cpp and a bunch of optimizations and yeah much much better. 128k context instead of 64k and significantly faster than stock ollama. It fixed in 5 minutes what it couldn't do in a day.\n\nStill need to try Gemma though.",
"sig": "2e6b7cec8fc08517e9708939f394f43510da076ac10c9d41ba6492b469efec6b2d60f8096e48b76fd69cab8e98349d79362a12e431ab3324c3eb9de4f16353ee"
}