Recently, MiniMax, a tech company, quietly changed its billing model from a per-task basis (Coding Plan) to a token-based billing model (Token Plan) when launching its new flagship model, MiniMax-M3. Due to the lack of prior communication with users and the significantly higher token consumption compared to expectations, many heavy users quickly exhausted their monthly quotas, leading to strong dissatisfaction and concentrated complaints from the developer community.

Facing public pressure, MiniMax released an apology statement on the evening of June 2nd, acknowledging that the adjustment lacked sufficient communication, the transition plan was not well considered, and the handling of old users' weekly limits was inappropriate. The company explained that the new M3 model is larger in size, supports native multimodal capabilities, and has a context length of up to one million tokens, requiring more computing resources. Additionally, switching to a token-based billing system is to meet the demand for users to freely use their subscription quotas across different modalities.

To compensate for the mistake and reward subscribed users, MiniMax announced the following compensation and benefit adjustment plans:

  1. Retention of Old User Benefits: For old users who purchased before March 22, 2026, and had the privilege of no weekly limit, their M2.7 and M3 models will continue to maintain the no weekly limit privilege after the upgrade.
  2. Permanent Weekly Limit Increase: For users who purchased the Token Plan before 10:00 on June 5, the weekly limit for the M3 model will be permanently increased by 50% during their valid subscription period.
  3. Temporary Quota Doubling: The company reset user quotas uniformly. From June 1 to June 7, all subscribed users' "5 hours/week" usage quota will temporarily double.
  4. Extended Points and Refund Channel: The validity period of previously issued compensation points has been automatically extended from 1 month to 1 year. Meanwhile, the online self-service refund channel is expected to go live on June 3 (Tuesday), allowing users to apply for refunds directly.

This incident is not an isolated case. Previously, Moonshot (Kimi) also faced user protests due to similar billing model adjustments. Industry experts point out that as large models evolve towards AI agent (Agent) scenarios, the token consumption per task has grown exponentially. Transitioning to a token-based billing model is an inevitable choice for companies to control computing costs. However, during the commercial transformation process, ensuring users' right to information and building a complete consumer trust system remain key challenges for the entire industry.