Skip to main contentLLMTune publishes upcoming functionality so teams can plan ahead. Timelines may shift, but the sequence reflects the current product strategy.
Recently Shipped
- FineTune Studio – No-code fine-tuning with guided workflows, real-time monitoring, and support for all training methods (SFT, DPO, PPO, RLAIF, CTO) and modalities
- LLMTune Models – Production-ready model catalog with comparison tools and deployment notes
- LLMTune API – OpenAI-compatible inference endpoints with streaming support and usage telemetry
- LLMTune Deploy – Deployment control with version management, traffic routing, and instant rollback
- LLMTune Evaluate – Comprehensive evaluation suite with automated scorecards and human review workflows
- Dataset Hub – Data intelligence platform with multiple sources, quality scoring, and PII detection
- Training Queue System – Sequential job processing to conserve GPU resources and ensure stability
- Federated & Traditional Compute – Flexible compute options (Single Instance or GPU Cluster) for all training methods
- All Modalities Support – Text, image, audio, video, code, multimodal, embeddings, and TTS
In Development
- Enhanced Dataset Hub – Direct HuggingFace Hub integration, S3/GCS connectors, and advanced data blending
- Improved Usage Dashboards – Deeper analytics for per-endpoint latency, error breakdowns, and budget alerts
- Advanced Traffic Management – More sophisticated canary and shadow deployment patterns
- Video Understanding Training – Full support for video-language model fine-tuning (currently in preview)
- Enhanced Evaluation Suite – More automated evaluation metrics and integration with training pipeline
Planned
- Team-level Permissions – Granular roles for dataset annotators versus deployment operators
- Secrets Management – Store third-party API keys used by custom inference workflows
- Self-hosted Agents – Export fine-tuned models with runtime configuration for on-prem orchestration
- Advanced Model Catalog – More models, better filtering, and community-contributed models
- Multi-workspace Management – Easier switching and management across multiple workspaces
- Enhanced Webhooks – More event types and better webhook management UI
Feature Requests
To request features or share feedback:
