Structured-output index

The same model breaks your pipeline's structured outputs 15× more often on some inference providers than on others. Find the best one for your pipelines.

Give this to your agent ↓

providers.dottxt.ai/llms.txt

← All providers

8 models · 3.5M structured outputs a day

#	Model	Broken structured outputs	Rank	Best for this model	Structured outputs a day
1	GLM 5.3 Flash	0.40%	3 of 29	Parasail (fp4) 0.28%	2.0M
2	Kimi K3	0.59%	3 of 20	Parasail (fp4) 0.20%	13k
3	GLM 5.3spot	0.62%	3 of 28	Together 0.25%	339k
4	GLM 5.3	0.73%	4 of 28	Together 0.25%	14k
5	DeepSeek V4.1 Flash 1017	0.91%	8 of 33	Reka AI 0.40%	20k
6	Kimi K3spot	0.94%	7 of 20	Parasail (fp4) 0.20%	75k
7	GLM 5.3fast	1.11%	8 of 28	Together 0.25%	27k
8	DeepSeek V4.1 Flash	1.19%	12 of 28	DeepInfra (fp8/flex) 0.35%	30k
9	DeepSeek V4.1 Flash 1017spot	1.20%	17 of 33	Reka AI 0.40%	268k
10	MiniMax M3.1 Flash	1.50%	7 of 8	Nebius Token Factory (fp4) 0.14%	3.4k
11	Kimi K3fast	1.59%	9 of 20	Parasail (fp4) 0.20%	4.9k
12	DeepSeek V4 Flash 0423	1.68%	5 of 13	Alibaba Cloud Int. (fp8) 0.66%	39k
13	DeepSeek V4.1 Flashspot	1.92%	15 of 28	DeepInfra (fp8/flex) 0.35%	704k
14	Nemotron 3.5 Lightning	2.83%	3 of 5	CoreWeave (bf16) 1.75%	2.3k

New models

Get updates when a new model drops

And which provider to use.

For inference providers

Structured outputs don't have to break

The same provider can handle structured outputs well on one model and struggle on the next. Implementations diverge on strings, optionals, composition and conditionals (the April announcement): the schema one provider enforces, another ignores.

Worst model2.83%

of Nemotron 3.5 Lightning's structured outputs fail on Fireworks.

Best model0.40%

on GLM 5.3 Flash, its best model.

Best on0 of 8

models, where Fireworks is the most reliable provider.

dotjson plugs into your inference stack, so every model you serve gets perfect structured outputs from day 0.

Fix Fireworks's structured outputs↵