[The operating cost of Kimi K2 Thinking is an order of magnitude lower than GPT-5] Jinshi Data reported on November 8 that despite Kimi K2 Thinking's parameter scale reaching trillions, its operating cost remains very low. Its API pricing is $0.15 per million tokens input (cache hit) / $0.6 per million tokens input (cache miss), and $2.5 per million tokens output, which is an order of magnitude lower than GPT-5's pricing ($1.25 per million tokens input, $10 per million tokens output).