of GLM 5.3'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 | Amazon Bedrock# | 0.08% | 13k | $1.68 / $5.28 | 84 | 80.5% |
| 2 | Friendli# | 0.08% | 3.4k | $1.26 / $3.96 | 96 | 100.0% |
| 3 | inference.netfp4# | 0.11% | 1.0M | $0.08 / $3.60 | 114 | 100.0% |
| 4 | SiliconFlowfp8# | 0.15% | 399k | $0.70 / $2.20 | 53 | 99.9% |
| 5 | ModelRun [by Modular]fp4# | 0.17% | 4.9k | $2.10 / $6.60 | 106 | 93.1% |
| 6 | Inceptronfp4# | 0.22% | 101k | $1.20 / $2.20 | 95 | 99.8% |
| 7 | Sail Researchfp8# | 0.28% | 111k | $0.50 / $2.00 | 49 | 100.0% |
| 8 | Waferfp4# | 0.30% | 1.5M | $0.07 / $1.40 | 96 | 99.8% |
| 9 | Modal# | 0.32% | 186k | $1.40 / $4.40 | 170 | 99.9% |
| 10 | Together# | 0.33% | 82k | $1.40 / $4.40 | 87 | 100.0% |
| 11 | Decartfp4# | 0.38% | 185k | $0.84 / $2.69 | 92 | 100.0% |
| 12 | Prime Intellect# | 0.38% | 57k | $0.98 / $3.08 | 68 | 99.9% |
| 13 | Morphfp8# | 0.41% | 21k | $0.18 / $1.54 | 48 | 98.7% |
| 14 | Reka AI# | 0.56% | 12k | $0.14 / $2.10 | 54 | 98.9% |
| 15 | Relace# | 0.57% | 84k | $0.07 / $4.38 | 77 | 100.0% |
| 16 | Basetenfp4# | 0.59% | 111k | $1.40 / $4.40 | 78 | 99.2% |
| 17 | Z.aifp8# | 0.65% | 830k | $1.40 / $4.40 | 57 | 100.0% |
| 18 | CoreWeavefp8# | 0.66% | 16k | $1.20 / $4.00 | 103 | 97.0% |
| 19 | Mistralnvfp4/zdr# | 0.70% | 52k | $1.40 / $4.40 | 106 | 99.9% |
| 20 | Nebius Token Factoryfp4# | 0.75% | 3.7k | $1.40 / $4.40 | 101 | 98.6% |
| 21 | Basetenfp8/fast# | 0.76% | 38k | $2.10 / $6.60 | 128 | 98.2% |
| 22 | Parasailfp8# | 0.79% | 28k | $1.40 / $4.40 | 130 | 98.3% |
| 23 | Makorafp4# | 0.79% | 32k | $0.14 / $3.00 | 67 | 96.7% |
| 24 | io.netfp8# | 0.86% | 2.9k | $0.33 / $3.20 | 75 | 99.9% |
| 25 | DekaLLM# | 0.90% | 13k | $0.25 / $2.95 | 66 | 99.9% |
| 26 | DeepInfrafp4# | 0.99% | 42k | $0.56 / $2.50 | 71 | 99.4% |
| 27 | Venice# | 1.17% | 2.1k | $1.40 / $4.40 | 88 | 96.9% |
| 28 | Fireworks# | 1.27% | 11k | $1.40 / $4.40 | 97 | 99.5% |
| 29 | Mistralnvfp4# | 1.40% | 9.1k | $1.40 / $4.40 | 134 | 100.0% |
| 30 | Crusoefp4# | 1.60% | 110k | $1.40 / $4.40 | 98 | 100.0% |
| 31 | Fireworksfast# | 1.67% | 81k | $2.10 / $6.60 | 133 | 99.2% |
| 32 | Alibaba Cloud Int.# | 1.82% | 10k | $1.19 / $3.74 | 48 | 100.0% |
| 33 | Fireworks Spotspot# | 1.87% | 178k | $0.98 / $3.08 | — | 99.9% |
| 34 | AkashMLfp8# | 2.25% | 9.2k | $0.19 / $4.40 | 78 | 100.0% |
| 35 | Cloudflare# | 2.70% | 6.2k | $1.40 / $4.40 | 35 | 100.0% |
| 36 | Tenstorrentfp8# | 2.77% | 24k | $0.60 / $2.50 | 68 | 99.2% |
| 37 | Alibaba Cloud Int.prime# | 2.85% | 7.7k | $2.80 / $8.80 | 107 | 100.0% |
| 38 | DigitalOcean# | 3.27% | 25k | $0.91 / $2.86 | 67 | 99.9% |
| 39 | Baidu Qianfanfp8# | 4.04% | 13k | $0.28 / $0.88 | 78 | 100.0% |
| 40 | Phala# | 4.43% | 7.3k | $0.84 / $2.64 | 42 | 100.0% |
| 41 | NovitaAIfp8# | 6.71% | 41k | $0.63 / $1.98 | 64 | 100.0% |
| 42 | GMICloudfp8# | 23.02% | 2.6k | $0.70 / $2.20 | 24 | 99.9% |
Also ignored, under 1,000 tool calls a day: AtlasCloud (fp8) 5.00%.
Recommended OpenRouter settings for GLM 5.3
{
"provider": {
"order": [
"amazon-bedrock/us-east-1",
"friendli",
"inference-net/fp4",
"inference-net/us"
],
"ignore": [
"digitalocean",
"baidu/fp8",
"phala",
"atlas-cloud/fp8",
"novita/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 Amazon Bedrock, its best provider.
broken tool calls a day across its 37 providers.