GLM 5.3 API Cost: Always Thinking, 2.5x Cheaper per Answer Than 5.2

GLM 5.3 costs the same $1.40 per million input tokens and $4.40 per million output as GLM 5.2, and it no longer lets you turn thinking off: every way of disabling it returns a 400 error, and the defau...