OpenAI Launches GPT Mini Latest with 400,000 Token Context Window

OpenAI released GPT Mini Latest on April 27, 2025, a new model featuring a 400,000 token context window. The model is designed to automatically redirect to the latest version in the OpenAI GPT Mini family.

Key Specifications

The model offers a 400,000 token context window, significantly larger than many competing models. Pricing per 1 million tokens has not yet been disclosed by OpenAI.

According to OpenRouter, the model functions as a pointer that always redirects to the newest GPT Mini family member, eliminating the need for developers to manually update their integrations when new versions are released.

Technical Capabilities

OpenRouter documentation indicates the model supports reasoning-enabled features, allowing it to display step-by-step thinking processes. Developers can enable this through the reasoning parameter in API requests and access the model's internal reasoning via the reasoning_details array in responses.

The model maintains conversation context by preserving complete reasoning_details when passing messages back, enabling continued reasoning from previous conversation points.

Availability

GPT Mini Latest is currently available through OpenRouter's API. OpenAI has not published benchmark scores, parameter count, or detailed training information at the time of release.

The model is positioned as part of OpenAI's GPT Mini family, though specific differentiators from other GPT Mini versions remain undisclosed.

What This Means

The auto-redirect feature addresses a common developer pain point: keeping applications updated with the latest model versions. By pointing to GPT Mini Latest, developers can ensure they're always using the newest capabilities without code changes. However, this approach may introduce unexpected behavior changes when OpenAI releases new versions, potentially requiring careful monitoring in production environments. The 400,000 token context window positions it competitively for long-document processing and extended conversations, though actual performance benchmarks are needed to assess its capabilities against alternatives from Anthropic, Google, and others.