Azure DeepSeek-V4-Pro and DeepSeek-V4-Flash#2067
Conversation
| reasoning = true | ||
| temperature = true | ||
| knowledge = "2025-05" | ||
| tool_call = true |
There was a problem hiding this comment.
this PR is very inaccurate, these models DONT support tool calls, you aren't using base_model syntax, and theese models are only served over completions api as far as i can tell and that isnt specified here either
There was a problem hiding this comment.
My apologies, you're right! I thought they did support tool calls, but it turns out they don't on Azure Foundry! https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure?tabs=global-standard&pivots=azure-direct-others#deepseek-models-sold-by-azure
I based this on other models within the provider, sorry, I wasn't aware of the base_model syntax. I've read the README now (again, my apologies) and it should be good.
The only thing I'm not sure about is the completions API bit - so to be safe I've copied that bit from Azure/Kimi K2.6's definition. I was actually expecting the completions API to be be: "https://${AZURE_RESOURCE_NAME}.services.ai.azure.com/openai/v1/chat/completions" but that's also the same for Kimi - so I went with what I know works already. Let me know if I got any of that wrong and I'll correct.
2 participants
Add DeepSeek-V4-Pro and DeepSeek-V4-Flash models to the Azure provider.
DeepSeek V4 model pricing on Azure can be found here: https://azure.microsoft.com/en-us/pricing/details/ai-foundry-models/deepseek/
All other details taken from Azure Foundry model pages.