2026 Changelog

Version 4.8

Compass Portal

API Key Limit Increase to 300 per Project
The maximum number of API keys that can be created per project has been increased to 300.

Compass API

GPT-5.2 (Preview)
One of the flagship models for coding and agentic tasks with configurable reasoning support.

GPT-4.1 mini
Optimized model for instruction following and tool calling, offering a 1M-token context window and low-latency responses without a dedicated reasoning step. The model is deployed in the UAE region.

Feb 05th 2026
Version 4.7.2

Compass API

GPT 5.1 Enhancements
GPT-5.1 is now available in production and in the UAE region.

Compass Playground

Enhanced API Key Management
The Playground now displays associated API keys sorted by most recently updated or created. Users without an associated API key will see a guided message explaining how to create and link one. This update improves usability and onboarding for both new and existing users.

Compass Portal

Support Ticket Submission from Compass Portal
Compass users can now submit support tickets directly from the navigation menu within the platform, enabling faster assistance without disrupting their workflow.

Compass API

GPT 5.1 (Preview)
An advanced model with enhanced understanding, reasoning ability, and designed to be more accurate, flexible, and efficient.

Single API Endpoint to Retrieve Model Definitions
Compass now provides a single, standardized API endpoint that allows users to retrieve all available model definitions within the platform.

API Key Visibility Controls
A new field, API Key Visibility, has been added in the Create API Key screen, allowing root admins to manage which department users can view, edit, or regenerate specific API keys rather than granting visibility to all users by default.

GPT-4o Model Routing 
Starting January 20, 2026, the Compass API model routing has been updated to standardize usage on a single GPT-4o version. Requests previously routed to gpt-4o, gpt-4o-2024-11-20, gpt-4o-2024-08-06, will now be routed exclusively to gpt-4o-2024-11-20.

Model Retirement
The GPT-4o model version gpt-4o-2024-08-06 has been retired. All requests sent to this model will now be automatically routed to gpt-4o-2024-11-20.

Model Deprecation and Retirement Guide
To help you plan as models evolve, Compass now includes a dedicated documentation page that clearly communicates model deprecations and retirements. The documentation provides information about the:

• Models that are already deprecated

• Models that are scheduled for deprecation

• Models that are retired or scheduled for retirement

Version 4.6.6

Compass API

LFM2-8B-A1B
A high-efficiency 8.3B parameter hybrid Mixture-of-Experts (MoE) language model designed for fast, memory-efficient inference on phones, laptops, and embedded systems, with support for both CPU- and GPU-based deployments. The model is deployed on the Core42 cloud in the UAE region.