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 default effort is max . On 11 tasks with a checkable answer, run three times each, max was the only setting that got all 33 right, at $0.00468 per correct answer, 2.5x less than GLM 5.2 at max because…
GLM 5.3 and its smaller variant, GLM 5.3 Flash, are retrained versions of the previous GLM 5.2 model. These models are priced the same as their predecessor at $1.40 per million input tokens and $4.40 per million output tokens, but they now have no option to turn off "thinking". Every setting defaults to "max", which means the model always reasons through the problem before answering.
In benchmark tests, GLM 5.3 achieved a 100% correct answer rate at $0.00468 per correct response, which is 2.5 times cheaper than GLM 5.2's maximum pricing of $0.01173 per correct answer. The smaller Flash model performed slightly worse, scoring 31 out of 33 correct answers at a tenth of the cost of GLM 5.3. This is due to its smaller parameter size, with only 18 out of 320 billion parameters being active during inference.
Unlike previous versions, GLM 5.3 does not allow the user to disable reasoning. When "thinking" is set to 'disabled', the API returns a 400 error. Instead, users must choose between low, high, or max for the reasoning_effort parameter. At the default max setting, GLM 5.3 scored 33 out of 33 correct answers, while GLM 5.3 Flash achieved 31 correct answers at a much lower price.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
