Open Models Almanac
Methodsinference & servingagentic scaffolding

general family · filed under inference & serving

Harmony format

A structured conversation format required for correct use of the described gpt-oss models.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See agentic scaffolding for the methods filed beside it.

Also called harmony response format.

sources
3
model
1
lab adopt it
1
strongest
core

Documented in

Evidence

3 quotes from 3 sources, one card per model, each opening with its strongest.

How sources treat it
core 3

gpt-oss-120b

OpenAIadopted3 sources

Both models were trained on our harmony response format and should only be used with the harmony format as it will not work correctly otherwise.

coreunclearin gpt-oss-120bOpenAI
2 more from this model

gpt-oss should not be used without using the harmony format, as it will not work correctly.

coreinference servingin gpt-ossOpenAI

should only be used with this format; otherwise, they will not work correctly.

coreinference servingin gpt-ossOpenAI
↑