Fine-tuning status (roadmap)
Meridian supports pass-through inference when you supply a pre-existing Azure OpenAI fine-tuned deployment ID. No uploads, no training runs — just point us at your endpoint.
What works today
- ✓Pass-through to any Azure OpenAI fine-tuned deployment via deployment ID and API key.
- ✓Streaming and non-streaming completions with your custom model.
- ✓Token usage tracking and cost attribution per request.
On the roadmap
- →Managed fine-tuning jobs — upload JSONL, we handle the Azure training run.
- →Built-in eval harness with pass@k and side-by-side comparisons.
- →One-click deploy from checkpoint to production endpoint.
Have a fine-tuned deployment ready? Quickstart guide →