of Qwen3.8 2.4T A95B's tool calls fail on Together.
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 | Qwen3.5-9B | 0.25% | 2 of 5 | Darkbloom (fp4) 0.08% | 149k |
| 2 | GLM 5.3 | 0.33% | 10 of 42 | Amazon Bedrock 0.08% | 82k |
| 3 | Kimi K3 | 0.36% | 20 of 24 | DeepInfra (mxfp4) 0.01% | 83k |
| 4 | Muse Spark Open | 0.38% | 3 of 6 | Nebius Token Factory (fp4) 0.07% | 34k |
| 5 | Nemotron 3.5 Super VL | 0.66% | 1 of 8 | — | 280k |
| 6 | GLM 5.2 | 0.76% | 11 of 27 | Friendli 0.15% | 60k |
| 7 | Beam 1.0 501B A23B | 1.01% | 1 of 3 | — | 15k |
| 8 | MiniMax M3.1 Flash | 1.17% | 1 of 11 | — | 76k |
| 9 | DeepSeek V4.1 Flash | 1.52% | 31 of 35 | Prime Intellect 0.06% | 2.1M |
| 10 | MiniMax M3 | 1.56% | 6 of 11 | ModelRun [by Modular] (fp8) 0.27% | 44k |
| 11 | GLM 5.3 Flash | 1.78% | 26 of 34 | NextBit (fp4) 0.00% | 3.0M |
| 12 | DeepSeek V4 Pro 0813 | 2.26% | 18 of 21 | Ionstream 0.04% | 42k |
| 13 | gpt-oss-120b | 2.81% | 5 of 15 | Groq 0.04% | 4.0k |
| 14 | DeepSeek V4.1 Flash 1017 | 3.44% | 32 of 35 | Sail Research (fp4) 0.03% | 5.8M |
| 15 | Qwen3.8 2.4T A95B | 4.19% | 5 of 5 | SiliconFlow (fp8) 0.19% | 12k |
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 Qwen3.5-9B, its best model.
models, where the settings tell OpenRouter to skip Together.