DeepSeek releases V4.1-Flash-1017 with open weights

News17 Oct 2026By Lawrence Ladomery

DeepSeek has upgraded V4.1 Flash. The API model name and prices stay the same.

On 17 October 2026, DeepSeek announced DeepSeek-V4.1-Flash-1017, an updated snapshot of its V4.1 Flash model. DeepSeek describes it as "a Pareto improvement in performance and cost" over the previous version.

According to DeepSeek's API changelog, developers keep using the deepseek-flash model name, and API usage and pricing are unchanged. DeepSeek says the weights are available to download from Hugging Face and ModelScope.

For benchmarks and full details, read DeepSeek's announcement.

Sources

DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made Real

Related model families

Family	Developer	Weights	Licence	Models
DeepSeek	DeepSeek	Open	MIT	3

Hosts that can deploy the models

Host	LLM hosting plans	Model families
Amazon AWS	Amazon EC2, Amazon Bedrock, Amazon SageMaker	Mistral, DeepSeek, Gemma, gpt-oss, Llama, Qwen
Google Cloud	Model Garden, Google Kubernetes Engine (GKE), Compute Engine	Llama, Mistral, DeepSeek, Qwen, gpt-oss, Gemma
Microsoft Azure	Azure Machine Learning, Azure Kubernetes Service (AKS), Azure GPU Virtual Machines	gpt-oss, Qwen, DeepSeek, Llama, Gemma, Mistral