According to Perception Beating monitoring, upon opening the MiniMax Open Platform's latest billing document, the price of the flagship model MiniMax-M3 has been marked with a red "Permanent 50% Discount." Under the standard pay-as-you-go mode, the price for one million tokens for inputs within 512k tokens has been directly slashed from $0.60 to $0.30, and for outputs from $2.40 to $1.20. Even for inputs exceeding 512k tokens with extra-long context, the price has been halved accordingly.
The significant price reduction follows a severe trust crisis that occurred less than half a month ago. On June 1st, when MiniMax introduced the M3 model, it abruptly switched from its original pay-per-call mode to a token-based billing, effectively reducing the subscription benefits of existing users. This sudden change caused many developers to experience a sudden cost increase of over 250%. Faced with widespread criticism on platforms like V2EX, and users flocking to complaint platforms and consumer rights associations, the parent company, RareSky Technology, was forced to swiftly issue an apology on the evening of June 2nd, resetting user quotas and providing additional compensation.
However, the compensation announcement did not completely halt the outflow of developers. Amid the fierce price war in the domestic large model market, especially with competitors like DeepSeek continuously poaching talent at extremely low prices, MiniMax ultimately chose to compromise entirely on June 15th, just two weeks after the release of the M3 model, by permanently halving its price. The drastic turnaround from "quietly raising prices" to "slashing prices for survival" reflects the passivity of startups in pricing and commercialization strategies.
While the halving of prices temporarily stabilized developers, it sparked chain concerns in the secondary capital market. Goldman Sachs pointed out that the aggressive price cut would severely erode profit margins, leading to a 14% target price reduction for MiniMax. JPMorgan also lowered its rating, stating directly that a rapid price reduction after a new model release often signals that the model's actual capabilities are below expectations. In this trade-off between buying into the intelligence spectrum and shorting MiniMax in the Hong Kong stock market, investors have significantly favored the opposing side.
