Remy Reference/models/DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731

DeepSeek V4 Flash 0731, a language model from DeepInfra.

modelOverride: { model: "deepseek-v4-flash-0731" }

DeepSeek·served via

DeepInfra·Released August 2026

Pricing

Input: $0.090 / 1M
Output: $0.18 / 1M

Specifications

Context window: 1,048,576 tokens
Max output: 384,000 tokens
Released: August 2026
Model string: deepseek-ai/DeepSeek-V4-Flash-0731
Capabilities: Reasoning

Parameters

Model options, passed in modelOverride.config.

Parameter	Type	Default	Range / options
`reasoning_effort` Non-think for fast responses, High for complex problem-solving, Max to push reasoning to its fullest extent.	Select	high	noneminimallowmediumhighxhigh+1
`top_p` Nucleus sampling. Considers only tokens whose cumulative probability exceeds this threshold.	Number	0.9	0–1, step 0.01
`top_k` Limits sampling to the K most likely tokens at each step. Set to 0 to disable.	Number	0	0–100
`min_p` Minimum probability threshold relative to the most likely token.	Number	0	0–1, step 0.01
`presence_penalty` Penalizes tokens that have already appeared in the output, encouraging new topics.	Number	0	-2–2, step 0.01
`frequency_penalty` Penalizes tokens based on how often they have already appeared.	Number	0	-2–2, step 0.01
`repetition_penalty` Penalizes repeated tokens. Values above 1 discourage repetition.	Number	1.05	0–2, step 0.01
`seed`	Seed	—	-1–2147483647

Use this model

await mindstudio.generateText({
  message: "...",
  modelOverride: {
    model: "deepseek-v4-flash-0731",
    config: {
      reasoning_effort: "high",
      top_p: 0.9,
    },
  },
});