GLM 5.3 Forces Mandatory Thinking, Cuts Cost Per Correct Answer by 60%

Z.ai has released GLM 5.3, a post-trained upgrade to GLM 5.2 that shares identical pricing at $1.40 per million input tokens and $4.40 per million output tokens. Unlike its predecessor, GLM 5.3 makes reasoning mandatory — attempts to disable thinking return an HTTP 400 error, and only low, high, and max effort levels are supported. In benchmark testing across 11 verifiable tasks run three times each, GLM 5.3 at maximum reasoning effort answered all 33 correctly at $0.00468 per correct answer, roughly 2.5 times cheaper than GLM 5.2 at the same setting because the new model reasons about half as much. A smaller companion model, GLM 5.3 Flash, achieved 31 of 33 correct answers at approximately one-tenth the cost, featuring 320 billion total parameters with only 18 billion activated per token. Both models offer a 1 million-token context window and show significant gains over GLM 5.2 on agent, coding, and automation benchmarks.