of gpt-oss-120b's tool calls fail on Crusoe (bf16).
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 0423fp8 | 0.46% | 1 of 16 | — | 159k |
| 2 | DeepSeek V4.1 Flashfp8 | 0.75% | 21 of 35 | Prime Intellect 0.06% | 135k |
| 3 | Kimi K2.6bf16 | 0.86% | 7 of 14 | SiliconFlow (fp8) 0.06% | 15k |
| 4 | DeepSeek V4 Pro 0423fp8 | 1.18% | 3 of 13 | Cloudflare 0.78% | 4.6k |
| 5 | DeepSeek V4.1 Flash 1017fp8 | 1.46% | 25 of 35 | Sail Research (fp4) 0.03% | 313k |
| 6 | GLM 5.3fp4 | 1.60% | 30 of 42 | Amazon Bedrock 0.08% | 110k |
| 7 | GLM 5.3 Flashfp4 | 1.65% | 25 of 34 | NextBit (fp4) 0.00% | 368k |
| 8 | Gemma 4 31Bbf16 | 1.75% | 7 of 12 | ModelRun [by Modular] (fp4) 0.09% | 23k |
| 9 | Nemotron 3.5 Super VLbf16 | 4.58% | 5 of 8 | Together 0.66% | 8.4k |
| 10 | gpt-oss-120bbf16 | 7.33% | 9 of 15 | Groq 0.04% | 15k |
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 0423, its best model.
models, where the settings tell OpenRouter to skip Crusoe.