Meta Muse Wonder beats Opus 5.5 and GPT-6 Astra on agent tasks

Meta's Muse Wonder outperformed Opus 5.5 and GPT-6 Astra on SWE-Marathon, AutomationBench and MemoryBench, though Opus still leads the overall AA index.

K2S

@k2sbhaiArchived 15 days agoOriginally posted 2026.10.13

Meta Muse Wonder beats Opus 5.5 and GPT-6 Astra on agent tasks (image 1 of 4)

Meta Muse Wonder beats Opus 5.5 and GPT-6 Astra on agent tasks (image 2 of 4)

Meta Muse Wonder beats Opus 5.5 and GPT-6 Astra on agent tasks (image 3 of 4)

Meta Muse Wonder beats Opus 5.5 and GPT-6 Astra on agent tasks (image 4 of 4)

K2S @k2sbhaiOriginally posted 2026.10.13

Open original on X

Meta's Muse Wonder just beat Opus 5.5 and GPT-6 Astra on agent work 😳 - SWE-Marathon: 57.5 vs 46.3 opus - AutomationBench: 45.4 vs 41.6 gpt-6 - MemoryBench: 85.5 vs 81.5 opus > overall AA index still has opus on top (58 vs 56)

x.com/k2sbhai/status/2109692394266619943Archived 2026.10.13 10:12

Open original on X