of DeepSeek V4 Flash 0731's tool calls fail on Alibaba Cloud Int.
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 | Sail Researchfp4# | 0.04% | 95k | $0.15 / $0.30 | 41 | 100.0% |
| 2 | Wafernvfp4/fast# | 0.24% | 676k | $0.09 / $0.23 | 150 | 100.0% |
| 3 | Relacefp4# | 0.41% | 882k | $0.03 / $0.19 | 90 | 99.3% |
| 4 | Inceptronfp4# | 0.56% | 23k | $0.13 / $0.38 | 63 | 100.0% |
| 5 | CoreWeavefp8# | 1.01% | 423k | $0.13 / $0.28 | 134 | 100.0% |
| 6 | AtlasCloudfp4# | 1.17% | 1.1k | $0.44 / $1.32 | 77 | 99.9% |
| 7 | NovitaAIfp8# | 1.27% | 106k | $0.41 / $1.23 | 65 | 100.0% |
| 8 | Cloudflare# | 1.38% | 12k | $0.44 / $1.32 | 71 | 99.9% |
| 9 | GMICloudfp8# | 1.45% | 76k | $0.29 / $0.86 | 62 | 100.0% |
| 10 | DigitalOcean# | 1.61% | 1.0M | $0.12 / $0.24 | 166 | 100.0% |
| 11 | Venice# | 1.79% | 41k | $0.13 / $0.26 | 61 | 99.9% |
| 12 | Cohere# | 1.82% | 192k | $0.14 / $0.28 | 178 | 99.8% |
| 13 | Parasailfp8# | 1.84% | 135k | $0.14 / $0.28 | 88 | 99.9% |
| 14 | StreamLakefp8# | 2.09% | 455k | $0.13 / $0.40 | 43 | 100.0% |
| 15 | Baidu Qianfanfp8# | 2.09% | 754k | $0.05 / $0.14 | 155 | 99.5% |
| 16 | Basetenfp8# | 2.31% | 103k | $0.13 / $0.26 | 134 | 99.9% |
| 17 | Basetenfp32# | 2.65% | 36k | $0.13 / $0.26 | 94 | 99.9% |
| 18 | SiliconFlowfp8# | 2.94% | 4.1k | $0.22 / $0.66 | 31 | 100.0% |
| 19 | DeepInfrafp8# | 4.03% | 587k | $0.06 / $0.18 | 46 | 100.0% |
| 20 | Phala# | 7.10% | 9.8k | $0.31 / $0.92 | 93 | 100.0% |
| 21 | Reka AI# | 10.07% | 225k | $0.10 / $0.28 | 147 | 99.7% |
| 22 | Alibaba Cloud Int.# | 15.01% | 24k | $0.18 / $0.53 | 82 | 98.7% |
Also ignored, under 1,000 tool calls a day: Mancer (fp8) 6.93%.
{
"provider": {
"order": [
"sail-research/fp4",
"sail-research/us",
"wafer/fast",
"relace/fp4"
],
"ignore": [
"deepinfra/fp8",
"mancer/fp8",
"phala",
"reka",
"alibaba"
]
}
}Point your agent at llms.txt
And which provider to use.
Releasing a model? Test it before launch.
of DeepSeek V4 Flash 0731's tool calls fail on Alibaba Cloud Int.
as often on Alibaba Cloud Int. as on Sail Research (fp4), its best provider.
broken tool calls a day across its 21 providers.