# DeepSeek-V4.1-Flash (1017)

## Key facts

- Model author: DeepSeek
- Price: from $0.30 per 1M input tokens and $1.20 per 1M output tokens (tiered; this is the starting tier)
- Context length: 1M tokens
- Maximum output: 384K tokens
- Input modalities: Text, Image
- Output modalities: Text
- Capabilities: Reasoning, Tool use, Vision
- Release date: 2026-10-19
- Model IDs: `deepseek/deepseek-v4.1-flash-1017`
- Model page: https://www.realrelay.ai/model/deepseek/deepseek-v4.1-flash-1017
- API base URL: https://www.realrelay.ai/v1 (OpenAI-compatible)

## How to call it

RealRelay.ai serves DeepSeek-V4.1-Flash (1017) through one OpenAI-compatible API. Send a RealRelay.ai API key and set the `model` field to `deepseek/deepseek-v4.1-flash-1017`.

## Supported parameters

Request parameters the model author's own channel accepted in test calls on 2026-10-19. Other channels may differ.

- Core: `messages`, `messages[].role=system`, `model`
- Output length: `max_tokens`
- Sampling: `frequency_penalty`, `logprobs`, `presence_penalty`, `seed`, `temperature`, `top_k`, `top_logprobs`, `top_p`
- Stop sequences: `stop`
- Reasoning: `reasoning_effort`, `thinking.type`
- Structured output: `response_format.type=json_object`
- Tools: `parallel_tool_calls`, `tool_choice`, `tools`, `tools[].type=function`
- Streaming: `stream`, `stream_options.include_usage`