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.
| # | Model | Broken tool calls | Rank | Best on the model | Tool calls a day |
|---|---|---|---|---|---|
| 1 | DeepSeek V4 Flash Vision Expfp8 | 0.58% | 1 of 3 | — | 5.0k |
| 2 | GLM 5fp8 | 0.74% | 1 of 5 | — | 3.8k |
| 3 | DeepSeek V4 Flash 0731fp8 | 1.45% | 9 of 22 | Sail Research (fp4) 0.04% | 76k |
| 4 | DeepSeek V3.2fp8 | 1.61% | 4 of 4 | SiliconFlow (fp8) 0.60% | 32k |
| 5 | DeepSeek V4 Pro 0423fp8 | 1.63% | 5 of 13 | Cloudflare 0.78% | 6.2k |
| 6 | Hy3bf16 | 1.67% | 3 of 5 | Tencent Cloud (fp8) 1.46% | 7.5k |
| 7 | DeepSeek V4.1 Flashfp8 | 1.89% | 32 of 35 | Prime Intellect 0.06% | 30k |
| 8 | DeepSeek V4 Flash 0423fp8 | 1.98% | 9 of 16 | Crusoe (fp8) 0.46% | 684k |
| 9 | MiMo-V2.6-Flashbf16 | 2.79% | 5 of 9 | Venice (fp8) 1.41% | 29k |
| 10 | MiniMax M2.7fp8 | 3.22% | 5 of 6 | Groq 0.32% | 7.4k |
| 11 | MiMo-V2.6-Probf16 | 3.77% | 2 of 4 | DeepInfra (fp8) 1.48% | 227k |
| 12 | MiniMax M3.1 Flashfp8 | 3.88% | 10 of 11 | Together 1.17% | 47k |
| 13 | DeepSeek V4 Pro 0813fp8 | 4.21% | 20 of 21 | Ionstream 0.04% | 3.5k |
| 14 | MiniMax M3fp8 | 4.87% | 10 of 11 | ModelRun [by Modular] (fp8) 0.27% | 21k |
| 15 | DeepSeek V4.1 Flash 1017fp8 | 6.88% | 35 of 35 | Sail Research (fp4) 0.03% | 60k |
| 16 | GLM 5.3 Flashfp8 | 7.78% | 34 of 34 | NextBit (fp4) 0.00% | 140k |
| 17 | GLM 5.3fp8 | 23.02% | 42 of 42 | Amazon Bedrock 0.08% | 2.6k |
Get updates when a new model drops
And which provider to use.
For inference providers
Tool calling doesn't have to break
The same provider can handle tool calls well on one model and struggle on the next.
on DeepSeek V4 Flash Vision Exp, its best model.
models, where the settings tell OpenRouter to skip GMICloud.