DeepSeek price bumps: GLM-5.3 is stronger and peaks are cheaper

DeepSeek had just forecast a substantial increase in prices, while the GLM-5.3 of the spectrometer had just caught up. Assuming GLM-5.3 follows the API price of GLM-5.2, the average token price for the two peak-period models has been roughly flat. The GLM-5.3 input of $8 per million token, output of $28 and V4-pro price of $9 and $27. The GLM-5.3 performance was stronger in most of the same-calibre assessments, especially in the context of the need for consecutive code writing, call tools and long missions. DeepSeek remains cheap in the low valleys and in the Cache period, with v4-pro inputting and output down to 4.5 and 13.5 yuan in the low valleys, and as low as 0.15 in the Cache, compared to $2 in the GLM. GLM-5.3 is more competitive when it runs during the day。