of MiniMax M3's tool calls fail on AtlasCloud (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 Pro 0813fp8 | 0.30% | 3 of 21 | Ionstream 0.04% | 2.9k |
| 2 | Hy4 previewfp8 | 0.35% | 5 of 6 | DeepInfra (fp4) 0.10% | 13k |
| 3 | DeepSeek V4.1 Flashfp8 | 0.92% | 25 of 35 | Prime Intellect 0.06% | 166k |
| 4 | Kimi K2.6int4 | 0.98% | 8 of 14 | SiliconFlow (fp8) 0.06% | 1.9k |
| 5 | DeepSeek V4 Flash 0731fp4 | 1.17% | 6 of 22 | Sail Research (fp4) 0.04% | 1.1k |
| 6 | Kimi K2.5int4 | 1.25% | 2 of 5 | SiliconFlow (int4) 0.17% | 3.0k |
| 7 | Qwen3.5 397B A17Bfp8 | 1.55% | 5 of 7 | Alibaba Cloud Int. 0.58% | 1.6k |
| 8 | Hy3fp8 | 1.63% | 2 of 5 | Tencent Cloud (fp8) 1.46% | 3.9k |
| 9 | MiniMax M2.7fp8 | 1.93% | 3 of 6 | Groq 0.32% | 2.5k |
| 10 | MiMo-V2.5-Profp8 | 2.04% | 1 of 2 | — | 2.9k |
| 11 | GLM 5.3 Flashfp8 | 2.15% | 28 of 34 | NextBit (fp4) 0.00% | 27k |
| 12 | Qwen3.6 35B A3Bfp8 | 2.46% | 4 of 6 | AkashML (fp8) 1.63% | 9.3k |
| 13 | DeepSeek V4.1 Flash 1017fp8 | 3.67% | 34 of 35 | Sail Research (fp4) 0.03% | 231k |
| 14 | DeepSeek V4 Pro 0423fp4 | 3.78% | 10 of 13 | Cloudflare 0.78% | 2.2k |
| 15 | DeepSeek V4 Flash 0423fp4 | 4.27% | 14 of 16 | Crusoe (fp8) 0.46% | 5.2k |
| 16 | MiniMax M3fp8 | 36.08% | 11 of 11 | ModelRun [by Modular] (fp8) 0.27% | 16k |
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 Pro 0813, its best model.
models, where the settings tell OpenRouter to skip AtlasCloud.