DeepSeek’s Price Surge Fades into Obscurity? Flash's Peak Price Once Surpassed GPT-5.6 Luna

According to Sentinel Beating monitoring, after a sharp increase in price, the DeepSeek V4 API's cost-effectiveness has surprisingly not been undermined. Starting from August 17, during the V4-Flash peak, the price surged to $3 per million tokens for input and $9 per million tokens for output, making the output price 4.5 times higher than before. However, a similar-tier model, GPT-5.6 Luna, has inputs of $1 and outputs of $6. Even at DeepSeek's highest peak, the Flash input is still about 56% cheaper, with the output price only about 22% of Luna's.

Compared to other budget models, Flash has not lost its advantage. The GLM-5.2 model, also priced at 51 points, has median input and output prices of $1.4 and $4.4 per million tokens, which is still about 3 times higher than the post-surge Flash prices. MiniMax M3 is even cheaper, with inputs at $0.3 and outputs at $1.2 per million tokens, but with an IQ of only 44 points. During DeepSeek's idle period, Flash is priced at around $0.22 and $0.67, making it even more affordable.

The situation for V4-Pro is not as exaggerated. During its peak, it is priced at around $1.33 for input and $4 for output, which is already comparable to Muse Spark 1.2 at $1.25/$4.25, but still lower than Grok 4.6 at $2/$6.