of GLM 5.3 Flash's tool calls fail on GMICloud (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 | NextBitfp4# | 0.00% | 584k | $0.12 / $0.48 | 98 | 99.8% |
| 2 | Friendli# | 0.10% | 268k | $0.15 / $0.50 | 81 | 100.0% |
| 3 | inference.netfp4# | 0.12% | 3.2M | $0.11 / $0.35 | 98 | 100.0% |
| 4 | Reka AI# | 0.18% | 255k | $0.12 / $0.30 | 117 | 100.0% |
| 5 | Modalfp4# | 0.20% | 150k | $0.15 / $0.50 | 96 | 100.0% |
| 6 | Particle AInvfp4# | 0.23% | 27k | $0.15 / $0.50 | 124 | 98.6% |
| 7 | Wafernvfp4# | 0.25% | 930k | $0.05 / $0.20 | 135 | 99.0% |
| 8 | DeepInfrafp4# | 0.27% | 228k | $0.15 / $0.50 | 5 | 100.0% |
| 9 | Sail Researchfp4# | 0.36% | 61k | $0.08 / $0.25 | 20 | 99.8% |
| 10 | Nebius Token Factoryfp4# | 0.38% | 12k | $0.15 / $0.50 | 108 | 84.9% |
| 11 | Morphfp8# | 0.44% | 209k | $0.14 / $0.23 | 38 | 99.7% |
| 12 | Prime Intellect# | 0.49% | 171k | $0.15 / $0.50 | 46 | 100.0% |
| 13 | Inceptronfp8# | 0.55% | 179k | $0.17 / $0.60 | 79 | 99.9% |
| 14 | Cohere# | 0.71% | 1.1M | $0.15 / $0.50 | 122 | 98.3% |
| 15 | Phalanvfp4# | 0.89% | 612k | $0.13 / $0.42 | 43 | 99.2% |
| 16 | NEAR AIfp8# | 0.94% | 479k | $0.08 / $0.50 | 23 | 99.7% |
| 17 | DigitalOcean# | 0.97% | 1.3M | $0.11 / $0.38 | 96 | 100.0% |
| 18 | Basetenfp8# | 1.25% | 626k | $0.15 / $0.50 | 76 | 98.9% |
| 19 | Makora# | 1.25% | 275k | $0.14 / $0.47 | 86 | 99.5% |
| 20 | Open Inferencefp4# | 1.46% | 1.5M | $0.05 / $0.13 | 90 | 99.9% |
| 21 | Venice# | 1.57% | 56k | $0.15 / $0.50 | 105 | 99.2% |
| 22 | Parasailfp4# | 1.58% | 65k | $0.15 / $0.50 | 126 | 99.8% |
| 23 | StreamLakefp8# | 1.59% | 94k | $0.09 / $0.29 | 45 | 95.6% |
| 24 | Z.aifp8# | 1.63% | 1.1M | $0.15 / $0.50 | 38 | 100.0% |
| 25 | Crusoefp4# | 1.65% | 368k | $0.15 / $0.50 | 94 | 99.5% |
| 26 | Together# | 1.78% | 3.0M | $0.15 / $0.50 | 76 | 96.7% |
| 27 | Relace# | 1.93% | 710k | $0.05 / $0.07 | 111 | 100.0% |
| 28 | AtlasCloudfp8# | 2.15% | 27k | $0.15 / $0.50 | 47 | 96.9% |
| 29 | CoreWeavenvfp4# | 2.18% | 1.5M | $0.12 / $0.38 | 147 | 99.1% |
| 30 | NovitaAIfp8# | 2.26% | 404k | $0.08 / $0.28 | 43 | 99.9% |
| 31 | Mistral# | 2.45% | 19k | $0.15 / $0.50 | 114 | 98.9% |
| 32 | Fireworks# | 4.20% | 625k | $0.15 / $0.50 | 94 | 99.2% |
| 33 | SiliconFlowfp8# | 4.60% | 34k | $0.15 / $0.50 | 31 | 95.4% |
| 34 | GMICloudfp8# | 7.78% | 140k | $0.07 / $0.25 | 83 | 96.3% |
Recommended OpenRouter settings for GLM 5.3 Flash
{
"provider": {
"order": [
"nextbit/fp4",
"friendli",
"inference-net/fp4",
"inference-net/us"
],
"ignore": [
"fireworks",
"siliconflow/fp8",
"gmicloud/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 GMICloud (fp8) as on NextBit (fp4), its best provider.
broken tool calls a day across its 34 providers.