conv.

← The whole story
AI
Open reproduction confirms 5-agent Qwen3.8-27B swarm can beat Fable 5 on coding

Memdoor publishes full parity benchmark beating Fable 5

1 Oct 31 6:46 PM · 11d ago · 1 article · 1 post · 2 sources · development 1 of 4

The rebuilt harness, run on OpenRouter with the authors' exact prompts, scored a mean 9.8/10 over five passes on ten LiveCodeBench Hard problems versus 9.6 for the original harness and 8.0 for Fable 5 alone; all 50 runs, costs and hosts were published, with 49 of 50 problems solved.

Memdoor publishes full parity benchmark beating Fable 5
github.com

“49 of 50 solved; the 50 runs cost $35.79 (15.3M output tokens), $38.77 with the 16 attempts that were cut by a clock, stopped for a fix or paused.”

guregodevo (Memdoor README)

guregodevo Author/submitter of the Memdoor reproductionLee Author of the original GVS5H paper (2026)Fable 5 Proprietary model used as the comparison baselineQwen3.8-27B Open-weight model used in the five-agent team

The whole story articlespostscomments the bright band is this development · numbered dots are the others · click one to jump

Peak 14 pieces in 3h at Oct 31, 6 PM; 40 pieces over 11 days (2 articles · 3 posts · 35 comments) Oct 31, 6 PM — 14 pieces · 2 articles · 2 posts · 10 comments — Hacker News 11, Newswires 2, Mastodon 1Oct 31, 9 PM — 4 pieces · 1 post · 3 comments — Hacker News 3, Mastodon 1Nov 1, 12 AM — 5 pieces · 5 comments — Hacker News 5Nov 1, 2 AM — 2 pieces · 2 comments — Hacker News 2Nov 1, 5 AM — 3 pieces · 3 comments — Hacker News 3Nov 1, 8 AM — 5 pieces · 5 comments — Hacker News 5Nov 1, 11 AM — 4 pieces · 4 comments — Hacker News 4Nov 1, 2 PM — 2 pieces · 2 comments — Hacker News 2Nov 1, 5 PM — quietNov 1, 8 PM — quietNov 1, 11 PM — quietNov 2, 2 AM — quietNov 2, 5 AM — quietNov 2, 8 AM — quietNov 2, 11 AM — 1 piece · 1 comment — Hacker News 1Nov 2, 2 PM — quietNov 2, 5 PM — quietNov 2, 8 PM — quietNov 2, 11 PM — quietNov 3, 2 AM — quietNov 3, 5 AM — quietNov 3, 8 AM — quietNov 3, 11 AM — quietNov 3, 2 PM — quietNov 3, 5 PM — quietNov 3, 8 PM — quietNov 3, 11 PM — quietNov 4, 2 AM — quietNov 4, 5 AM — quietNov 4, 8 AM — quietNov 4, 11 AM — quietNov 4, 2 PM — quietNov 4, 5 PM — quietNov 4, 8 PM — quietNov 4, 11 PM — quietNov 5, 2 AM — quietNov 5, 5 AM — quietNov 5, 8 AM — quietNov 5, 11 AM — quietNov 5, 2 PM — quietNov 5, 5 PM — quietNov 5, 8 PM — quietNov 5, 11 PM — quietNov 6, 2 AM — quietNov 6, 5 AM — quietNov 6, 8 AM — quietNov 6, 11 AM — quietNov 6, 2 PM — quietNov 6, 5 PM — quietNov 6, 8 PM — quietNov 6, 11 PM — quietNov 7, 2 AM — quietNov 7, 5 AM — quietNov 7, 8 AM — quietNov 7, 11 AM — quietNov 7, 2 PM — quietNov 7, 5 PM — quietNov 7, 8 PM — quietNov 7, 11 PM — quietNov 8, 2 AM — quietNov 8, 5 AM — quietNov 8, 8 AM — quietNov 8, 11 AM — quietNov 8, 2 PM — quietNov 8, 5 PM — quietNov 8, 8 PM — quietNov 8, 11 PM — quietNov 9, 2 AM — quietNov 9, 5 AM — quietNov 9, 8 AM — quietNov 9, 11 AM — quietNov 9, 2 PM — quietNov 9, 5 PM — quietNov 9, 8 PM — quietNov 9, 11 PM — quietYesterday, 2 AM — quietYesterday, 5 AM — quietYesterday, 8 AM — quietYesterday, 11 AM — quietYesterday, 2 PM — quietYesterday, 5 PM — quietYesterday, 8 PM — quietYesterday, 11 PM — quietToday, 2 AM — quietToday, 5 AM — quietToday, 8 AM — quietToday, 11 AM — quietToday, 2 PM — quietToday, 5 PM — quietToday, 8 PM — quiet 1–4
1–4Memdoor publishes full parity benchmark…
Nov 1Nov 2Nov 3Nov 4Nov 5Nov 6Nov 7Nov 8Nov 9yesterdaynow · 10:40 PM ET

Reported in the same hours no headline names this development itself — these 1 claim were published in its stretch

  1. 2 outlets Memdoor: A team of 5 Qwen3.8-27B models beats Fable 5 on LiveCodeBench Hard

    first by HN Best, 11d ago · also HN Frontpage

All 4 developments of Open reproduction confirms 5-agent Qwen3.8-27B swarm can… →

Hacker NewsMastodonNewswires