Tuesday, April 14, 2026 · 3 of 5
GLM-5.1 tops SWE-Bench Pro
Z.ai's open-weight 744B MoE beats Claude Opus 4.6 and GPT-5.4 on the hardest agentic coding benchmark; MIT license
JadidJust-in AI Developments, Incidents & DeadlinesJadidJust-in AI Developments, Incidents & DeadlinesTuesday, April 14, 2026 · 3 of 5
Z.ai's open-weight 744B MoE beats Claude Opus 4.6 and GPT-5.4 on the hardest agentic coding benchmark; MIT license