DeepSeek V4.1 Flash 1017 Arrives on Workers AI with a 1M-Token Context and Image Input!

Hi everyone, it's Shii-chan! I found a new model on Workers AI today and I'm super excited to share it!

Cloudflare Changelog developers.cloudflare.com

What was announced?

The Cloudflare Changelog announced that @cf/deepseek-ai/deepseek-v4.1-flash-1017 is now available on Workers AI!

DeepSeek V4.1 Flash is a multimodal Mixture-of-Experts (MoE) model that can handle images too. It has 552B backbone parameters, and activates 8B parameters per token during prefill and 16B during decode. Cloudflare says this keeps input-heavy agentic workloads fast and low-cost.

Why it matters

The 1017 release is an update to the existing DeepSeek V4.1 Flash. The post describes improvements across reasoning, coding, agentic tasks, and visual benchmarks, and says it completes tasks with shorter outputs.

Shorter outputs can mean less waiting and fewer output tokens, which is nice if you run agents!

What changes

Workers AI now has a larger MoE model with long context and image input in its lineup. Since you can call it from Workers, it's easier to try things like feeding in long documents, adding image-based queries, or building agents that call tools.

Dive Deep

Key capabilities

• Vision: send images alongside text in the same request
• Reasoning: on by default at high effort. Set reasoning_effort to low or max to change it, or to none to turn reasoning off
• Function calling: build agents that invoke tools and APIs across multiple conversation turns
• Long context: a full 1,048,576-token context window

Requirements and rate limits

• Requires the Workers Paid plan or prepaid AI Gateway credits
• Default rate limit is 20 requests per minute with standard Workers AI billing
• 50 requests per minute with prepaid AI Gateway credits

How to call it

• Workers AI binding (env.AI.run())
• REST API
• OpenAI-compatible endpoint
• AI Gateway

The post doesn't list specific prices. It points to the Workers AI pricing page and the model page.

Wrap-up

• @cf/deepseek-ai/deepseek-v4.1-flash-1017 is available on Workers AI
• A 552B-backbone MoE that activates 8B (prefill) and 16B (decode) parameters per token
• Improved reasoning, coding, agentic, and visual benchmarks, with shorter outputs
• Supports image input, adjustable reasoning effort, function calling, and a 1,048,576-token context
• Needs Workers Paid or prepaid AI Gateway credits, with rate limits of 20 / 50 requests per minute

This one is for anyone who wants to run long-context agents on Workers AI!