Open-weight GLM-5.3 matches top model on coding tasks at half cost

Original: GLM-5.3 vs. GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing

Source: together.ai ↗