Mosaic AI Model Training
DatabricksMosaic AI Model Training was Databricks' managed service for full fine-tuning and continued pre-training of Meta Llama 3.1 to 3.3 models, driven from the databricks_genai SDK or a web UI. Training jobs read data from Unity Catalog and registered the resulting weights back to it, so customers kept the trained checkpoint. Databricks has ended support for the service and recommends its AI Runtime environment instead.
Databricks' own pages call this capability Foundation Model Fine-tuning, part of Databricks Model Training; neither page uses the name Mosaic AI Model Training.
Openness
1 high confidence- source
- closed
- training-code
- closed
- runs-on
- Databricks-workspace-only(AWS regions)
- license
- Proprietary(managed service)
- base-weights
- Llama-Community-License(non-OSI)
A proprietary managed fine-tuning service that runs only inside a Databricks workspace, with no source published. Databricks' separate LLM Foundry and Composer libraries are open source, but they are not this product.
- https://docs.databricks.com/aws/en/large-language-models/foundation-model-training/ recorded 2026-09-23
Foundation Model Fine-tuning requires a Databricks workspace in AWS us-east-1 or us-west-2 and is used only through the databricks_genai SDK or its UI; the page publishes no repository or self-hosted build of the training engine, and the only license text on the page covers the base Llama models being fine-tuned, not the service itself.
- https://www.databricks.com/product/machine-learning/mosaic-ai-training recorded 2026-09-09
Databricks markets this capability as Databricks Model Training, describing it only as a way to fine-tune an open source LLM on Databricks; the page offers no download, repository, or self-host option for the training service itself, and never uses the name Mosaic AI Model Training.
Adoption
not assessedDatabricks published no usage figure for Mosaic AI Model Training: no jobs run and no customer count. Nothing Databricks publishes about its wider platform is specific to this service, so the map leaves adoption unscored rather than borrow a figure from the platform as a whole.
- https://docs.databricks.com/aws/en/large-language-models/foundation-model-training/ recorded 2026-09-28
Searched the body of the Foundation Model Fine-tuning page for usage, job-count and customer figures: none found; the only status statement is that the feature "has reached end of life (EoL) and is no longer supported".
- https://www.databricks.com/product/machine-learning/mosaic-ai-training recorded 2026-09-28
Searched the body of the Databricks Model Training product page for customer counts, job counts and usage figures specific to this fine-tuning capability: none found; it carries only marketing copy such as "Fine-tune an open source LLM or build custom LLMs trained on your enterprise data with Databricks Model Training".
Capability
2 medium confidenceMosaic AI Model Training offered chat-completion and instruction fine-tuning and continued pre-training on Meta Llama models, with no preference or reinforcement-learning methods, no LoRA or QLoRA, and no parallelism, precision or scale controls. That left it behind the managed offerings from OpenAI, Vertex and Together, which combine supervised, preference and reinforcement tuning.
- https://docs.databricks.com/aws/en/large-language-models/foundation-model-training/ recorded 2026-09-28
The Foundation Model Fine-tuning page lists exactly three supported tasks, chat completion, instruction fine-tuning and continued pre-training ("Train your model with additional text data" for the last), and its supported models table lists only "meta-llama/Llama-3.3-70B-Instruct" and other Meta Llama 3.1 and 3.2 checkpoints; searched the body for LoRA, QLoRA, DPO, RLHF and preference tuning: none found. It also carries a warning that the feature "has reached end of life (EoL) and is no longer supported".
Verified 2026-09-23