Pax Historia initially offered players access to AI-powered gameplay through token-based usage. That worked well in the early phase, especially while experimenting with free access to DeepSeek models and validating user demand.

But as usage scaled, the limits of token-based pricing became clear.

Players did not want to think in tokens or variable usage costs. They wanted a fixed monthly price they could understand immediately. At the same time, Pax Historia needed to make sure any subscription model stayed aligned with the real cost of serving different models.

That created a difficult product and infrastructure problem:

• Players wanted predictable, fixed-price access.
• Different models had very different GPU cost profiles.
• A simple unlimited subscription would not be economically sustainable.
• The team needed a system that was easy for users to understand and easy to operate at scale.

We worked with Pax Historia to design and implement a subscription-based simple pass system that replaced token-based pricing with request-based access.

The initial setup introduced a pass for Mimo V2 Flash, with a limit of 300 requests per day. This provided a high-volume option for fast, frequent gameplay interactions and allowed users to switch to a simple monthly subscription instead of variable billing.

As the product evolved, GLM-5 became the primary model offering for new users due to its stronger performance. A second pass was introduced for GLM-5, with a limit of 150 requests per day, reflecting its higher GPU consumption.

Existing users who had already adopted the Mimo pass were allowed to continue using it, ensuring service continuity while the platform transitioned toward GLM-5 as the default option.

This approach gave Pax Historia a clear and flexible structure for AI access:

• GLM-5 pass for new users seeking a premium model experience
• Mimo V2 Flash pass maintained for existing high-volume users

Behind the scenes, we handled quota tracking and enforcement directly at the platform layer, allowing Pax Historia to roll out this transition smoothly without building custom usage infrastructure from scratch.

The subscription pass model created a better experience for both Pax Historia and its players, by incorporating the passes with Pax Historia patron-based system.

For players, the value proposition became much clearer. Instead of worrying about token spend, they could choose a monthly pass based on the kind of experience they wanted.

For Pax Historia, the business model became more predictable and more scalable. Daily request caps helped put a firm upper bound on usage, while model-specific passes aligned subscription value with actual infrastructure cost.

The result was a tighter fit between product, pricing, and platform:

• A simpler, more intuitive pricing experience for users
• Clear segmentation between different AI model tiers
• Better control over inference cost exposure
• A more sustainable path to scaling AI-native gameplay

Pax Historia is a strong example of a broader shift happening across AI-native applications.

Token-based pricing may work at the infrastructure layer, but end users rarely want to buy AI that way. They want straightforward products with clear value and predictable costs.

By turning model access into subscription passes, Pax Historia made AI easier to buy, easier to use, and easier to scale.

That is the real win: not just serving models, but packaging AI in a way that matches how users actually want to consume it.