Dear Black Forest Labs team,
I hope this message finds you well.
we have been an active user of BFL’s finetuning capability, particularly for achieving personalized art styles in our products. Over time, we have created nearly 500 fine-tuned models, which now form a core pillar of our service offering.
Given your recent announcement that the Finetuning API (and related endpoints) will be deprecated as of October 31, 2025 (as indicated in your Release Notes)
docs.bfl.ai
, I am deeply concerned about the potential impact on our business. To ensure we can plan accordingly, I would greatly appreciate your clarification on a few key points:
- Rationale Behind the Decision
Could you share the primary reasons and considerations (technical, product, security, cost, strategic, etc.) that led to the decision to discontinue support for finetuning?
Was this decision driven by internal constraints (infrastructure cost, model maintenance, abuse risk) or a shift in product / technology direction?
Did BFL conduct an internal evaluation of user impact / risk before arriving at this deprecation decision?
- Future of Model Customization / Fine-tuning Capabilities
Although the announcement states “finetuning functionality will be discontinued” with “no migration path available” (https://docs.bfl.ai/release-notes), is there any possibility that BFL will reintroduce a variant of fine-tuning (for example via adapters, LoRA, modular style layers, or plugin modules) in a future architecture?
Do you have plans to offer lighter, style-adaptation, or concept-tuning capabilities (rather than full fine-tuning) as an alternative?
If a new customization pathway is planned, would there be a migration or transition path for existing users?
- Support for Already Trained / Paid Models
For all finetuned models that we or other users have already created (and paid for) before October 31, 2025, will they remain usable for inference after that date?
If yes, are there any limitations (such as restricted API endpoints, deprecated interfaces, or region constraints)?
If they will no longer be supported, will BFL provide any compensation, transitional access, or migration assistance?
Will BFL shut down the inference endpoints for finetuned models, and if so, what is the schedule / timeline for such shutdown?
- Transition / Mitigation Strategy for Customers
For customers relying heavily on finetuned models (such as ourselves), does BFL plan to offer a grace period, legacy support program, or extended access plan to reduce disruption risk?
Is there a possibility of custom agreements or enterprise-level support to preserve continuity for critical customers?
Would BFL consider enabling export or transfer of model parameters / weights (under licensing / IP policy) to alternative platforms, where feasible?
Thank you very much for your time and attention. I look forward to your response and hope we can find a path forward that balances your technical roadmap and the needs of committed users like us.
Dear Black Forest Labs team,
I hope this message finds you well.
we have been an active user of BFL’s finetuning capability, particularly for achieving personalized art styles in our products. Over time, we have created nearly 500 fine-tuned models, which now form a core pillar of our service offering.
Given your recent announcement that the Finetuning API (and related endpoints) will be deprecated as of October 31, 2025 (as indicated in your Release Notes)
docs.bfl.ai
, I am deeply concerned about the potential impact on our business. To ensure we can plan accordingly, I would greatly appreciate your clarification on a few key points:
Could you share the primary reasons and considerations (technical, product, security, cost, strategic, etc.) that led to the decision to discontinue support for finetuning?
Was this decision driven by internal constraints (infrastructure cost, model maintenance, abuse risk) or a shift in product / technology direction?
Did BFL conduct an internal evaluation of user impact / risk before arriving at this deprecation decision?
Although the announcement states “finetuning functionality will be discontinued” with “no migration path available” (https://docs.bfl.ai/release-notes), is there any possibility that BFL will reintroduce a variant of fine-tuning (for example via adapters, LoRA, modular style layers, or plugin modules) in a future architecture?
Do you have plans to offer lighter, style-adaptation, or concept-tuning capabilities (rather than full fine-tuning) as an alternative?
If a new customization pathway is planned, would there be a migration or transition path for existing users?
For all finetuned models that we or other users have already created (and paid for) before October 31, 2025, will they remain usable for inference after that date?
If yes, are there any limitations (such as restricted API endpoints, deprecated interfaces, or region constraints)?
If they will no longer be supported, will BFL provide any compensation, transitional access, or migration assistance?
Will BFL shut down the inference endpoints for finetuned models, and if so, what is the schedule / timeline for such shutdown?
For customers relying heavily on finetuned models (such as ourselves), does BFL plan to offer a grace period, legacy support program, or extended access plan to reduce disruption risk?
Is there a possibility of custom agreements or enterprise-level support to preserve continuity for critical customers?
Would BFL consider enabling export or transfer of model parameters / weights (under licensing / IP policy) to alternative platforms, where feasible?
Thank you very much for your time and attention. I look forward to your response and hope we can find a path forward that balances your technical roadmap and the needs of committed users like us.