What it does
This entry is the Flash model on the DeepSeek API. The pricing page now calls it deepseek-flash, version DeepSeek-V4.1-Flash-1017, and says the older names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted. The models originally behind those names are retired, so requests under them are handled by the current Flash model and charged at the Flash rate. The model supports thinking mode, which is the default, and non-thinking mode. It has a 1M-token context and can return up to 384K tokens. Listed features include JSON output, tool calls, a Responses API, an Anthropic-format API, chat prefix completion (beta) and FIM completion (beta, non-thinking mode only). The page marks vision as supported for Flash but not for the Pro model beside it. The API is reachable at one OpenAI-format base URL and one Anthropic-format base URL, and Flash has a concurrency limit of 2,500. Billing is per million tokens and depends on the time of day. Peak hours are 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, excluding Chinese public holidays, and off-peak usage costs half as much as peak. Off-peak, Flash input costs $0.15 on a cache miss or $0.003 on a cache hit, and output costs $0.60. Fees come out of a topped-up or granted balance. Separate Search and Fetch APIs are billed at $0.75 per 1,000 successful requests.
Capabilities
- Thinking and non-thinking modes
- 1M-token context with up to 384K tokens of output
- JSON output, tool calls, Responses API and Anthropic-format API
- Vision support listed for Flash
- Off-peak usage at half the peak price
When to use it
- You want a low-cost hosted model with a very long context for agent or tool-calling workloads
- You already use OpenAI-format or Anthropic-format clients and want to point them at another endpoint
- You can run heavy batch work outside peak hours to halve the per-token rate
