Tool-call index

The same model breaks your agent's tool calls 18× more often on some inference providers than on others. Find the best one for your agents.

Give this to your agent ↓

StepFun · 2 providers · 4k tool calls a day

#	Provider		Tool calls a day	$ in / out per M	Tokens/s	Uptime
1	StepFunfp8#	1.28%	2.0k	$1.00 / $2.70	134	100.0%
2	Featherlessfp4#	4.47%	2.2k	$1.00 / $2.70	23	84.1%

Recommended OpenRouter settings for Step 5 Preview

{
  "provider": {
    "order": [
      "stepfun/fp8"
    ],
    "ignore": [
      "featherless/fp4"
    ]
  }
}

Give llms.txt to your agent

Using Step 5 Preview?

And which provider to use.

For model makers

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

Worst provider4.47%

of Step 5 Preview's tool calls fail on Featherless (fp4).

Best to worst4×

as often on Featherless (fp4) as on StepFun (fp8), its best provider.

Every day124

broken tool calls a day across its 2 providers.

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