of DeepSeek V4.1 Flash's tool calls fail on NovitaAI (fp8).
Updated October 29, 2026
Tool-call index
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 | Prime Intellect# | 0.06% | 139k | $0.30 / $1.20 | 130 | 99.9% |
| 2 | Morphfp8# | 0.13% | 883k | $0.04 / $0.38 | 108 | 99.8% |
| 3 | Relace# | 0.14% | 1.5M | $0.07 / $0.50 | 105 | 100.0% |
| 4 | Wafermxfp4# | 0.26% | 1.2M | $0.04 / $0.50 | 170 | 100.0% |
| 5 | Nebius Token Factoryfp8# | 0.28% | 35k | $0.30 / $1.20 | 48 | 99.3% |
| 6 | DeepInfrafp8/priority# | 0.30% | 104k | $0.21 / $0.63 | 79 | 52.7% |
| 7 | Fireworks Spotspot# | 0.31% | 1.2M | $0.21 / $0.84 | 65 | 100.0% |
| 8 | Open Inferencefp4# | 0.32% | 1.0M | $0.04 / $1.04 | 45 | 99.6% |
| 9 | inference.netfp8# | 0.33% | 1.5M | $0.10 / $0.60 | 92 | 100.0% |
| 10 | Decartfp4# | 0.34% | 407k | $0.09 / $0.18 | 56 | 97.0% |
| 11 | Sail Researchfp4# | 0.41% | 429k | $0.09 / $0.25 | 49 | 100.0% |
| 12 | Fireworks# | 0.46% | 224k | $0.30 / $1.20 | 197 | 99.6% |
| 13 | Baidu Qianfanfp8# | 0.50% | 975k | $0.03 / $0.13 | 81 | 100.0% |
| 14 | Google Vertex# | 0.50% | 99k | $0.30 / $1.20 | 63 | 98.9% |
| 15 | Reka AI# | 0.51% | 46k | $0.10 / $0.84 | 78 | 96.8% |
| 16 | CoreWeavefp8# | 0.55% | 1.1M | $0.13 / $0.65 | 210 | 99.5% |
| 17 | DigitalOcean# | 0.60% | 1.5M | $0.23 / $0.90 | 163 | 100.0% |
| 18 | DeepInfrafp8# | 0.64% | 2.2M | $0.14 / $0.42 | 67 | 100.0% |
| 19 | Parasailfp8# | 0.66% | 184k | $0.30 / $1.20 | 181 | 100.0% |
| 20 | Makorafp8# | 0.70% | 43k | $0.26 / $1.17 | 136 | 99.9% |
| 21 | Crusoefp8# | 0.75% | 135k | $0.29 / $1.20 | 67 | 99.3% |
| 22 | Alibaba Cloud Int.# | 0.79% | 160k | $0.17 / $0.66 | 121 | 100.0% |
| 23 | Cohere# | 0.88% | 110k | $0.30 / $1.20 | 120 | 99.7% |
| 24 | Venicefp8# | 0.89% | 203k | $0.30 / $1.20 | 126 | 99.9% |
| 25 | AtlasCloudfp8# | 0.92% | 166k | $0.14 / $0.54 | 107 | 99.8% |
| 26 | SiliconFlowfp8# | 1.04% | 68k | $0.30 / $1.20 | 114 | 100.0% |
| 27 | StreamLakefp8# | 1.24% | 103k | $0.15 / $0.59 | 70 | 100.0% |
| 28 | Basetenfast# | 1.26% | 26k | $0.60 / $2.40 | 186 | 87.5% |
| 29 | Modal# | 1.29% | 23k | $0.30 / $1.20 | 61 | 100.0% |
| 30 | Baseten# | 1.46% | 1.1M | $0.30 / $1.20 | 136 | 99.8% |
| 31 | Together# | 1.52% | 2.1M | $0.30 / $1.20 | 194 | 100.0% |
| 32 | GMICloudfp8# | 1.89% | 30k | $0.18 / $0.72 | 93 | 99.9% |
| 33 | Phala# | 2.74% | 32k | $0.21 / $0.84 | 143 | 99.8% |
| 34 | DeepInfrafp8/flex# | 3.20% | 13k | $0.11 / $0.34 | 79 | 97.6% |
| 35 | NovitaAIfp8# | 5.20% | 331k | $0.20 / $0.78 | 99 | 97.6% |
Recommended OpenRouter settings for DeepSeek V4.1 Flash
{
"provider": {
"order": [
"primeintellect",
"morph/fp8",
"relace"
],
"ignore": [
"deepinfra/fp8/flex",
"novita/fp8"
]
}
}Point your agent at llms.txt
Get updates when its successor drops
And which provider to use.
For model makers
Broken tool calls carry your model's name
Releasing a model? Test it before launch.
as often on NovitaAI (fp8) as on Prime Intellect, its best provider.
broken tool calls a day across its 31 providers.