Model components / Fine-tuning code

Azure OpenAI Fine-Tuning

Azure-managed fine-tuning of GPT-4o, GPT-4o-mini, GPT-4.1, and GPT-3.5 with enterprise controls (VNet, private endpoints, RBAC, content filters) that the raw OpenAI API doesn't expose. Picked by enterprises that already standardize on Azure for governance and data residency. Same underlying training methods as OpenAI's API (SFT, DPO) but billed through Azure and deployed to Azure-managed endpoints.

Azure OpenAI fine-tuning via Azure AI Foundry (docs live, GA, June 2026). Methods: SFT, DPO, RFT (reinforcement fine-tuning). Tunable: gpt-4o-mini/gpt-4o (SFT+DPO), gpt-4.1/-mini/-nano (SFT+DPO), o4-mini (RFT), gpt-5 (RFT, gated/invite). Preview: Ministral-3B, Qwen-32B, Llama-3.3-70B-Instruct, gpt-oss-20b (SFT). Proprietary managed Microsoft Azure service.

Openness

1 high confidence

1.0

source: closed
training-code: closed
runs-on: Azure-AI-Foundry-only
license: Proprietary(managed Azure service)

Proprietary managed FT service inside Azure AI Foundry; no source, runs only on Azure against (mostly) closed OpenAI models.

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/fine-tuning recorded 2026-06-04
managed FT service; SFT/DPO/RFT; gpt-4o/4.1/o4-mini/gpt-5 tunable; Azure-only

Adoption

not assessed

No disclosed standalone usage figure for the Azure OpenAI fine-tuning feature located this run (jobs run / customers tuning). Azure OpenAI overall is widely deployed in enterprise, but no honest signal specific to the FT SKU was found; declining to assign a level rather than borrow the umbrella surface number.

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/fine-tuning recorded 2026-06-04
GA managed FT feature; no usage figures disclosed

Capability

4 medium confidence

4.0

Strong managed method coverage (SFT+DPO+RFT) over frontier closed base models plus an open-model preview lane, but black-box: no parallelism/precision/scale controls, no MLPerf-Training basis. Scored 4 on method breadth + frontier base quality, calibrated equal to the OpenAI FT API (C4) it wraps; not 5 since the user cannot control training internals/scale.

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/fine-tuning recorded 2026-06-04
SFT, DPO, RFT across gpt-4o/4.1/o4-mini/gpt-5 + Qwen/Llama/gpt-oss preview

Unchanged since 2026-06-09 (last edited, not re-checked)