More Freedom with OpenAI-Compatible Endpoints

VTX Macro is broadening how traders can bring their own models into the platform. In addition to our existing direct providers and local options, this release adds a dedicated OpenAI-compatible endpoint flow designed for services that speak the OpenAI API format.

This opens the door to setups like Azure OpenAI, Together.ai, and other compatible endpoints, while keeping configuration inside the same profile-based system you already know. The new flow also remembers endpoint-specific models and credentials, making it much easier to switch between different custom backends without re-entering everything each time.

Smoother Local AI Experience

Local AI gets a meaningful usability upgrade as well. The system now does a better job of remembering your previously used local endpoints, restoring the matching model, and securely pulling the right saved key for that exact server.

Connection testing is more informative, model discovery is more flexible, and single-model endpoints can be applied with less manual setup. Whether you are running Ollama, LM Studio, llmster, or similar local stacks, you will experience a much smoother path from connection to active model selection.

WebGPU Enhancements

On the browser side, WebGPU setup is becoming more integrated with the rest of the platform. Model selection is more tightly connected to your main AI preferences, and the system does a better job of tracking downloaded browser models and preserving usable selections.

Across AI settings more broadly, model pickers are being organized more clearly across sources like direct providers, OpenAI-compatible endpoints, Local AI, WebGPU, OpenRouter, and Venice. You'll benefit from better context-window display and cleaner carryover of your selected model into the main trading workflow.

Clearer, Predictable Billing

This release also brings an important update to Bring Your Own Key (BYOK) economics. For users running the AI trader in server execution mode, the platform fee is shifting to a flat, predictable per-trader-call charge, replacing the previous multiplier-based structure for that path.

Insights and Trade Analysis calls are explicitly excluded from this fee. The result is a simpler, more predictable billing story for users who bring their own provider keys while still relying on VTX Macro for server-side trading automation.

Under the Hood

These updates also support a more reliable day-to-day experience. Profiles now carry more AI configuration directly, renamed or stale model IDs are repaired gracefully when your preferences are loaded, and the system and trade surfaces are better aligned around your currently selected model. For you, that translates into fewer mismatches between what was configured, what appears in the UI, and what the platform actually runs.

The net effect is straightforward: more freedom in where your models come from, less friction in managing them, and clearer behavior when those models are used in live trading workflows.