Memdoor: five Qwen3 models beat Fable 5 on LiveCodeBench
The gist
Memdoor's rebuilt five-agent workflow scored 9.8 of 10 on ten LiveCodeBench Hard problems, versus 8.0 for Fable 5 alone.
A recalculation calls the gap statistically insignificant, so the claimed win is a tie.
How it unfolded
- Mar 29, 2025 to Apr 6, 2025The LiveCodeBench Hard contests used in the Memdoor five-pass test date from this period.
- Oct 25, 2026The coder wrote the team-solve files itself from a description of the method; nobody hand-wrote them.
- Oct 31, 2026Memdoor's rebuilt GVS5H harness on Qwen3.8-27B scored a mean of 9.8 of 10 over five passes on the first ten LiveCodeBench Hard problems, against 9.6 for the authors' own harness and 8.0 for Fable 5 alone.