The short answer

1017 is the 17 October 2026 update to V4.1 Flash. Call it with deepseek-flash. The release notice keeps the calling method and prices unchanged, while reporting improvements in task completion and output efficiency.

FOR: Readers searching for DeepSeek 1017 and developers updating an API integration

What does 1017 mean?

The official change log dates DeepSeek-V4.1-Flash-1017 to 17 October 2026. It updates V4.1 Flash on the website, app and API. Search phrases such as DeepSeek 1017 and DeepSeek V4.1 Flash 1017 refer to this release.

DeepSeek reports improvements to coding and agent tasks, with shorter outputs and faster completion in its evaluations. Those published benchmark results depend on the harness and evaluation settings. Compare the model on a representative task from your own workflow before changing a production default.

Which API model name should I use?

Use deepseek-flash with the official API base URL https://api.deepseek.com. The release version is DeepSeek-V4.1-Flash-1017, while deepseek-flash is the documented API identifier. The current rate card also lists deepseek-v4-pro as a separate model.

The rate card states that the retired deepseek-v4-flash and deepseek-v4-flash-vision-exp names remain accepted as compatibility routes to V4.1 Flash 1017. Update new integrations to the documented deepseek-flash name so the intended model is clear.

TERMINAL
curl --fail-with-body https://api.deepseek.com/chat/completions \
  -H "Authorization: Bearer $DEEPSEEK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"deepseek-flash","messages":[{"role":"user","content":"Explain this function in three points."}]}'

Prices, context and image inputs

On the official rate card reviewed on 5 November 2026, Flash costs USD 0.15 per million uncached input tokens, USD 0.003 per million cache-hit input tokens and USD 0.60 per million output tokens during off-peak hours. Peak rates are twice these amounts. Estimate the full input and output workload, including conversation history and additional agent calls.

The same table lists a 1M-token context window and vision support for Flash. V4 Pro has no image-input support in that table. A context limit, output limit and tool-specific upload limit describe different parts of a request, so check the integration you actually use when preparing a large or multimodal task.

Updating Harness and checking search charges

The release notice recommends DeepSeek Harness 0.2.1-rc.1 or later for this model. Check the current official installation instructions, keep your model setting explicit and compare the result of one small task before expanding the workflow.

The same announcement opened Search and Fetch to API users. These services have separate per-request charges. Their server-side tool billing through Harness, Claude Code and Codex began at 16:00 UTC on 31 October 2026, which is 1 November in Beijing. Include those requests alongside model tokens when estimating an agent budget.

Sources

Start with the original documents. Models and prices change; check the official pages before using them.

01Official Flash 1017 release announcement02Official API change log03Current models and prices04Search and Fetch billing