of gpt-oss-120b's tool calls fail on Nebius Token Factory (fp4).
Not all models are served equally by all providers. On a typical model, the worst provider breaks tool calls 19× as often as the best. Find the best one for your agents.
| # | Provider | Broken tool calls | Tool calls a day | $ in / out per M | Tokens/s | Uptime |
|---|---|---|---|---|---|---|
| 1 | Groq# | 0.04% | 110k | $0.15 / $0.60 | 359 | 100.0% |
| 2 | DeepInfrabf16# | 1.46% | 66k | $0.04 / $0.17 | 33 | 99.3% |
| 3 | CoreWeavefp4# | 2.16% | 90k | $0.03 / $0.17 | 65 | 99.8% |
| 4 | SambaNova# | 2.53% | 2.1k | $0.14 / $0.95 | 378 | 94.6% |
| 5 | Together# | 2.81% | 4.0k | $0.15 / $0.60 | 72 | 98.3% |
| 6 | AkashMLbf16# | 3.10% | 13k | $0.03 / $0.19 | 45 | 100.0% |
| 7 | Parasailfp4# | 3.47% | 21k | $0.10 / $0.75 | 204 | 100.0% |
| 8 | DekaLLMbf16# | 3.94% | 48k | $0.03 / $0.18 | 63 | 99.9% |
| 9 | Crusoebf16# | 7.33% | 15k | $0.05 / $0.25 | 302 | 99.4% |
| 10 | Basetenfp4# | 9.87% | 6.2k | $0.10 / $0.50 | 165 | 99.5% |
| 11 | NEAR AIfp4# | 10.37% | 154k | $0.03 / $0.15 | 46 | 99.9% |
| 12 | Veniceint4# | 17.23% | 12k | $0.03 / $0.15 | 22 | 99.1% |
| 13 | NovitaAIfp4# | 29.28% | 13k | $0.05 / $0.25 | 124 | 98.9% |
| 14 | Cerebrasfp16# | 36.42% | 70k | $0.35 / $0.75 | 794 | 100.0% |
| 15 | Nebius Token Factoryfp4# | 80.12% | 3.9k | $0.15 / $0.60 | 295 | 99.7% |
Also ignored, under 1,000 tool calls a day: Mancer (fp8) 4.47%, Phala 9.08%, Sail Research (fp4) 48.75%.
{
"provider": {
"order": [
"groq",
"deepinfra/bf16",
"coreweave/fp4"
],
"ignore": [
"akashml/bf16",
"parasail/fp4",
"dekallm/bf16",
"mancer/fp8",
"crusoe/bf16",
"phala",
"baseten/fp4",
"near-ai/fp4",
"venice/int4",
"novita/fp4",
"cerebras/fp16",
"sail-research/fp4",
"nebius/fp4"
]
}
}Point your agent at llms.txt
And which provider to use.
Releasing a model? Test it before launch.
of gpt-oss-120b's tool calls fail on Nebius Token Factory (fp4).
as often on Nebius Token Factory (fp4) as on Groq, its best provider.
broken tool calls a day across its 15 providers.