Overview
Released on October 17, 2026, DeepSeek-V4.1-Flash-1017 is an update to DeepSeek V4.1 Flash, positioned for efficient reasoning, coding, visual understanding, and agentic tasks. It is available through DeepSeek’s website, app, and API; model weights are also available under the MIT license. (api-docs.deepseek.com)
Capabilities and uses
- Supports text and image understanding, thinking and non-thinking modes, JSON output, tool calls, Responses API, and Anthropic-compatible API access. Thinking mode is the default. (api-docs.deepseek.com)
- Uses reinforcement learning with self-feedback to improve end-to-end project development. DeepSeek demonstrates uses including web games and 3D modeling, and reports improvements in reasoning, coding, agentic, and visual benchmarks over the previous V4.1 Flash version. (api-docs.deepseek.com)
- Optimized for DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy. DeepSeek recommends Harness version 0.2.1-rc.1 or later. (api-docs.deepseek.com)
Limits and considerations
- API context length: 1M tokens; maximum output: 384K tokens. The documented concurrency limit is 2,500. (api-docs.deepseek.com)
- Fill-in-the-middle completion is a beta feature available only in non-thinking mode. Chat prefix completion is also beta. (api-docs.deepseek.com)
- The API ID
deepseek-flashpoints to the current Flash model and therefore does not guarantee a permanently pinned model version. Model outputs and generated projects should be reviewed and tested before use. (api-docs.deepseek.com)
Official API pricing
Prices below are for deepseek-flash, per 1 million tokens, checked on November 15, 2026. Input and output are billed separately. The CNY and USD listings are separate official regional listings, not currency conversions of a single quoted package. Prices may change. (api-docs.deepseek.com)
| Official listing | Period | Input, cache hit | Input, cache miss | Output |
|---|---|---|---|---|
| Mainland China, CNY | Off-peak | CNY 0.02 | CNY 1 | CNY 4 |
| Mainland China, CNY | Peak | CNY 0.04 | CNY 2 | CNY 8 |
| International, USD | Off-peak | USD 0.003 | USD 0.15 | USD 0.60 |
| International, USD | Peak | USD 0.006 | USD 0.30 | USD 1.20 |
Peak hours are Monday–Friday, excluding Chinese public holidays: 09:00–12:00 and 14:00–18:00 Beijing time, equivalent to 01:00–04:00 and 06:00–10:00 UTC. All other hours, weekends, and Chinese public holidays are off-peak. (api-docs.deepseek.com)
Search and web fetch are separate per-use services, not token prices: each costs CNY 5 per 1,000 calls on the Chinese listing or USD 0.75 per 1,000 calls on the international listing, whether called directly or through model tools. (api-docs.deepseek.com)