How to Use DeepSeek V4 Flash 0731 on Composite

At a glance

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model with 13B active parameters out of 284B total. DeepSeek describes this re-post-trained revision as suited for coding, reasoning and agent workflows. It reads text only, accepts up to 1,048,576 tokens of context, and can write replies of up to 131,072 tokens.

Where it fits

The model lists tool calling with parallel calls, structured output and a JSON response format, so it works as the model behind a coding assistant or an agent that calls functions. For roleplay it accepts the full set of samplers that chat presets use, including min P, top A, repetition penalty and frequency and presence penalties, so an existing SillyTavern preset carries over without errors.

Setting reasoning effort

Thinking is optional. Add an effort in square brackets to the model name to choose it per key or per chat: deepseek/deepseek-v4-flash-0731:free[low] for a short think, [high] or [max] for harder problems. Without a suffix, the request body's reasoning settings apply. Each call on the free ID counts as one request, so effort changes speed, not how many requests you have left.

Compared with DeepSeek V4.1 Flash

DeepSeek V4.1 Flash is also on the free list. Pick 0731 when you only send text. Pick V4.1 Flash when you want to send images, or when you need replies longer than 131,072 tokens, which is rare outside batch generation. Both draw from the same 80 free requests a day, so switching between them to compare costs the same as staying on one.

A note on long prompts

Agent tools and coding assistants often send large files with every message. The window is big enough for most repositories' working set, but a huge prompt still takes time to process. Trim what the tool sends when replies are slow to start.

Connect it to your chat app

1. Create an API key on the API Keys page.
2. In SillyTavern, ChubAI, Agnai or RisuAI, choose an OpenAI-compatible provider and set the base URL to https://composite.lucidity.sh/v1. JanitorAI wants the full URL: https://composite.lucidity.sh/v1/chat/completions.
3. Paste the key, then type deepseek/deepseek-v4-flash-0731:free as the model name.
4. Send a short message to check the connection before you load a long chat.

Step-by-step setup for each app is in the JanitorAI and SillyTavern guides, and the docs hub covers the others.

From code, it is a normal OpenAI-compatible request:

curl https://composite.lucidity.sh/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek/deepseek-v4-flash-0731:free", "messages": [{"role": "user", "content": "Hello"}]}'

Free access (as of October 21, 2026)

On October 21, 2026, this model is on Composite's free list as deepseek/deepseek-v4-flash-0731:free. The plain deepseek/deepseek-v4-flash-0731 stays paid. Every account shares 80 requests a day across all of the free models, separate from Synth's count, and the count resets at midnight UTC. Credits and plans do not raise it.

The free list changes. Check the free models on the Models page for the current set, and the features guide for the rules. A :free id that is no longer on the list returns a model-not-found error and charges nothing. When the 80 requests are used up, requests stop with an error until the reset; Composite never charges your credits to continue. You can switch to deepseek/deepseek-v4-flash-0731 on credits or a plan, or to Synth.

More

• Model page with live pricing and rankings
• Model modifiers, including reasoning effort suffixes
• Best LLMs for Roleplay in 2026
• All model guides