.TXT Build pipelines?→Serve models?→Release models?↓
Updated October 31, 2026 · How we measure

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 ↓
← All models
DeepSeek · 13 providers · 230k structured outputs a day
#ProviderStructured outputs
a day
$ in / out
per M
Tokens/sUptime
1Cloudflare#0.39%7.0k$1.32 / $3.963699.7%
2Nebius Token Factoryfp8#0.47%3.3k$1.32 / $3.967081.3%
3Ionstream#0.53%20k$0.37 / $2.9336100.0%
4Together#0.83%9.6k$1.32 / $3.9611999.4%
5DeepInfrafp8#1.09%13k$1.30 / $2.605299.8%
6DekaLLM#1.43%2.1k$0.88 / $3.804399.2%
7AtlasCloudfp8#1.48%6.6k$1.32 / $3.962499.7%
8Baidu Qianfanfp8#1.62%36k$0.28 / $0.8541100.0%
9Sail Researchfp4#1.81%62k$0.68 / $1.2049100.0%
10Wafermxfp4#2.02%46k$0.12 / $4.009699.7%
11Alibaba Cloud Int.#3.60%9.3k$0.58 / $1.742699.0%
12Parasailfp8#3.61%1.5k$1.32 / $3.963499.6%
13CoreWeavefp8#4.69%13k$1.31 / $3.9612899.1%

Also ignored, under 1,000 structured outputs a day: DigitalOcean 3.29%, Venice 5.32%.

Recommended OpenRouter settings for DeepSeek V4 Pro 0813
{
  "provider": {
    "order": [
      "cloudflare",
      "nebius/fp8",
      "ionstream"
    ],
    "ignore": [
      "digitalocean",
      "alibaba",
      "parasail/fp8",
      "coreweave/fp8",
      "venice"
    ]
  }
}

Give llms.txt to your agent

Using DeepSeek V4 Pro 0813?

Get updates when its successor drops

And which provider to use.

For model makers

Broken structured outputs carry your model's name

Your users don't see the provider. They see your model failing.

Worst provider4.69%

of DeepSeek V4 Pro 0813's structured outputs fail on CoreWeave (fp8).

Best to worst12×

as often on CoreWeave (fp8) as on Cloudflare, its best provider.

Every day4k

broken structured outputs a day across its 13 providers.

Releasing a model? Make your own API the one every provider is measured against.

Set the bar for your model↵