DeepSeek

Deepseek V4 Flash

LLM

DeepSeek V4 Flash is a fast language model optimized for efficient chat, reasoning, coding, tool calling, and structured text generation.

deepseek-v4-flash

PROVIDERSDeepSeekENDPOINTSopenaiCAPABILITIEStext

PRICINGDynamic Pricing

Prices vary by usage tier, request parameters or context conditions.

Tier	Input / 1M	Output / 1M	Cache Read / 1M
`base`	$0.098	$0.195	$0.02

Parameter	Type	Default	Range / values	Description
`messages` *	`array`	—	—	Conversation messages used as model input
`max_tokens`	`integer`	`4096`	1 ~ 384000	Maximum number of tokens to generate
`temperature`	`number`	`1`	0 ~ 2	Controls randomness; lower values make outputs more deterministic
`top_p`	`number`	`1`	0 ~ 1	Controls nucleus sampling for response diversity
`thinking`	`object`	—	enabled · disabled	Controls whether thinking mode is enabled
`reasoning_effort`	`enum`	`high`	high · max	Reasoning effort level for tasks that require deeper analysis
`stop`	`string \| array`	—	—	Optional stop sequence or sequences that end generation
`response_format`	`object`	—	json_object	Controls structured output format. Use {"type":"json_object"} for JSON object output.
`tools`	`array`	—	—	Tool definitions for function calling; tool type should be "function"
`tool_choice`	`string`	`auto`	none · auto · required	Controls whether the model calls tools. For this channel, use string values such as "none", "auto", or "required".
`stream`	`boolean`	`false`	—	Whether to stream partial responses