benchrAll capabilities Building software
Fine-tune a hosted OpenAI model
Closed to new users. The documentation states the fine-tuning platform is being wound down and is no longer accessible to new users, with existing users able to create jobs temporarily.
DocsDeprecatedTestNot tested by benchr
OpenAI API
Overview
- Documented methods were supervised fine-tuning, vision fine-tuning, direct preference optimisation and reinforcement fine-tuning.
- Documented base models were gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, gpt-4o and o4-mini - all 2025-generation checkpoints.
- This record is kept because a plan built on it needs a replacement, not a deleted page.
How it works
- 01If you are a new user, do not design around this. It is closed.
- 02If you have existing jobs, treat the window as temporary and plan the exit now.
- 03Move behaviour-shaping work to a Skill or a cached system prompt - both are cheaper and neither needs a training run.
- 04Move genuine weight-level customisation to an open-weight model you host yourself.
Limits
- No longer accessible to new users.
- The base model list never moved past the 2025 generation, so even where access exists it does not reach current frontier checkpoints.
- The documentation does not state a minimum example count.
Verification
benchr test runs
benchr has not run this capability. No test has been published.
benchr test runs →
Evidence
Sources
- OpenAI - Fine-tuning guide, carrying the wind-down notice and the base model list Tier 1 · September 1, 2026
Not stated by the source
- The exact date existing access ends
- Whether any replacement customisation path is planned
Source-check history
- September 1, 2026DocsDeprecatedLogged directly as Broken on first entry: the vendor's own fine-tuning guide states the platform is winding down and is closed to new users. Kept as a record so plans that depended on it have somewhere to land.