of Kimi K2.6's structured outputs fail on Inceptron (int4).
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.
| # | Provider | Structured outputs a day | $ in / out per M | Tokens/s | Uptime | |
|---|---|---|---|---|---|---|
| 1 | Parasailint4# | 1.53% | 11k | $0.75 / $3.50 | 61 | 99.9% |
| 2 | Cloudflare# | 1.70% | 1.5k | $0.95 / $4.00 | 32 | 99.9% |
| 3 | CoreWeavefp4# | 2.24% | 52k | $0.65 / $3.41 | 115 | 99.3% |
| 4 | Veniceint4# | 2.65% | 4.1k | $0.75 / $3.50 | 18 | 97.2% |
| 5 | Moonshot AI# | 2.66% | 10.0k | $0.95 / $4.00 | 31 | 99.3% |
| 6 | StreamLakefp8# | 2.76% | 3.6k | $0.60 / $2.52 | 47 | 96.9% |
| 7 | NovitaAI# | 3.17% | 2.9k | $0.80 / $3.40 | 32 | 100.0% |
| 8 | AtlasCloudint4# | 3.36% | 4.3k | $0.95 / $4.00 | 40 | 99.4% |
| 9 | Baidu Qianfanfp4# | 3.56% | 49k | $0.39 / $1.64 | 72 | 99.9% |
| 10 | DeepInfrafp4# | 3.64% | 25k | $0.75 / $3.50 | 18 | 99.7% |
| 11 | Crusoebf16# | 4.14% | 16k | $0.70 / $3.50 | 27 | 98.9% |
| 12 | Chutesint4# | 5.07% | 4.3k | $0.39 / $1.80 | 38 | 99.8% |
| 13 | Inceptronint4# | 5.21% | 9.0k | $0.68 / $3.50 | 78 | 99.9% |
Also ignored, under 1,000 structured outputs a day: SiliconFlow (fp8) 3.19%, Phala 3.86%.
{
"provider": {
"order": [
"parasail/int4",
"cloudflare",
"coreweave/fp4"
],
"ignore": [
"novita",
"siliconflow/fp8",
"atlas-cloud/int4",
"baidu/fp4",
"deepinfra/fp4",
"phala",
"crusoe/bf16",
"chutes/int4",
"inceptron/int4"
]
}
}Give llms.txt to your agent
And which provider to use.
Your users don't see the provider. They see your model failing.
of Kimi K2.6's structured outputs fail on Inceptron (int4).
as often on Inceptron (int4) as on Parasail (int4), its best provider.
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↵