of Llama 3.3 70B Instruct's tool calls fail on AkashML (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 | Kimi K3fp4 | 0.52% | 23 of 24 | DeepInfra (mxfp4) 0.01% | 1.8k |
| 2 | Qwen3.8 27Bfp8 | 1.00% | 12 of 21 | Darkbloom (fp4) 0.07% | 9.2k |
| 3 | Qwen3.6 35B A3Bfp8 | 1.63% | 1 of 6 | — | 66k |
| 4 | GLM 5.3fp8 | 2.25% | 34 of 42 | Amazon Bedrock 0.08% | 9.2k |
| 5 | gpt-oss-120bbf16 | 3.10% | 6 of 15 | Groq 0.04% | 13k |
| 6 | Gemma 4 26B A4Bfp4 | 8.70% | 11 of 11 | NextBit (bf16) 0.00% | 1.9k |
| 7 | gpt-oss-20bfp4 | 20.78% | 6 of 7 | Darkbloom (fp8) 0.13% | 6.2k |
| 8 | Llama 3.3 70B Instructfp8 | 32.75% | 5 of 5 | Groq 2.21% | 9.0k |
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 Kimi K3, its best model.
models, where the settings tell OpenRouter to skip AkashML.