Supported Models & Methods

The four methods in 2025–2026

OpenAI offers four fine-tuning methods. SFT is the workhorse; DPO and RFT are specialized; vision FT is for images.

Model	SFT	DPO	Vision FT	RFT
gpt-4.1 (2025-04-14)	✅	✅	—	—
gpt-4.1-mini (2025-04-14)	✅	✅	—	—
gpt-4.1-nano (2025-04-14)	✅	✅	—	—
gpt-4o (2024-08-06)	✅	✅	✅	—
gpt-4o-mini (2024-07-18)	✅	✅	—	—
o4-mini (2025-04-16)	—	—	—	✅
gpt-3.5-turbo	✅	—	—	—

SFT = Supervised Fine-Tuning, the default. DPO = Direct Preference Optimization, train with preferred/rejected pairs. RFT = Reinforcement Fine-Tuning, uses a grading rubric (reasoning models). Vision FT = images in training examples.

Pricing (per 1M tokens, 2025–2026)

Model	Training	Inference (in / out)
gpt-4.1-mini	$0.80	$0.80 / $3.20
gpt-4o-mini	$3.00	$0.30 / $1.20
gpt-4o	$25.00	$3.75 / $15.00
o4-mini (RFT)	$100/hr	$4.00 / $16.00

Best value for most projects: gpt-4.1-mini. Cheaper to train than gpt-4o-mini and a stronger base. Reach for gpt-4o only when you need vision FT, and for o4-mini only when you need reasoning RFT.

Code

List currently fine-tunable models·python

from openai import OpenAI
client = OpenAI()

# Discover what's currently fine-tunable
models = client.models.list()
for m in models.data:
    # Heuristic: fine-tunable models include 'fine-tuning' permissions
    # in their metadata; cleaner approach is to check the docs page.
    if any(k in m.id for k in ["gpt-4.1", "gpt-4o", "o4-mini", "gpt-3.5"]):
        print(m.id)

The four methods in 2025–2026

Pricing (per 1M tokens, 2025–2026)

Code

External links

Exercise

Progress

Comments 0