Has DeepSeek Lost Its Edge After Price Hike? Flash Now More Expensive Than GPT-5.6 Luna During Peak Hours

According to monitoring by Beating Insights, the recent price hike for DeepSeek V4 has finally erased its most exaggerated price advantage. Starting August 17, V4-Flash's peak-hour pricing has increased to 3 yuan per million tokens for input and 9 yuan for output. The issue is that OpenAI recently slashed the price of its comparable GPT-5.6 Luna by 80%, bringing its input cost down to just $0.20 and output to $1.20. Based on the current exchange rate, Luna's input costs approximately 1.35 yuan and output 8.1 yuan. In other words, during peak hours, DeepSeek Flash is not only no longer cheaper but is roughly twice as expensive for input and slightly more expensive for output compared to Luna. On the Artificial Analysis intelligence index, the two models score 50 and 51 points respectively, with performance nearly identical. DeepSeek used to rely on its pricing to leave models in this tier far behind, but now it can't even undercut Luna during peak hours. Only during off-peak hours does DeepSeek regain some of its cost-effectiveness. Flash's input price drops to 1.5 yuan, roughly on par with Luna, while output costs 4.5 yuan, still about 44% cheaper. When cache hits are high, DeepSeek's price advantage remains evident. V4-Pro has also entered a more standard pricing range. During peak hours, its costs translate to approximately $1.33 for input and $4 for output, making it very close to Meta Muse Spark, a flagship model in the low-price tier, though still cheaper than Grok 4.6's $2/$6 pricing. So, after this round of price hikes, DeepSeek is still relatively affordable, but the era of being "so cheap it has no rivals" is now a thing of the past. [Original Link]