DeepSeek R1 and the open chain of reasoning: why see the solution path

DeepSeek R1 is a reasoning model that returns an open chain of reasoning: the response shows not only the result but the whole path to it.

R1 is no longer in the Yenisei catalog. Our reasoning models are Kimi K2 Thinking (kimi-k2-thinking), DeepSeek V4 Pro (deepseek-pro), and GPT-OSS 120B (gpt-oss-120b); any of them can be asked to write out the solution path right in the answer.

Why see the solution path

A regular model is a black box: a question goes in, an answer comes out. If the answer is wrong, you don't know where exactly it took a wrong turn. Visible intermediate steps help in three cases:

• Prompt debugging. You can see at which step the model misunderstood the task. Often the problem is not the model but the wording, and the chain shows that immediately.
• Trust in high-stakes tasks. If the model draws a financial or legal conclusion, you can check the logic instead of taking its word for it.
• Teaching and explanations. A detailed walkthrough of a math or algorithm problem is useful in itself.

Example

curl https://api.yenisei.ru/v1/chat/completions \
  -H "Authorization: Bearer $YENISEI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2-thinking",
    "messages": [
      {"role": "system", "content": "First write out the solution step by step, then give the answer."},
      {"role": "user", "content": "A basket holds 3 red and 5 blue balls. We draw two without replacement. What is the probability that both are blue?"}
    ]
  }'

The response will contain not just "5/14" but the steps: 5/8 for the first ball, 4/7 for the second, then the product. An error in a derivation like this is easy to spot by eye.

Where the solution path is not needed

A written-out solution path means extra output tokens, which means time and money. For a short factual answer or a simple rephrasing it is unnecessary. If you don't need to check the logic, use DeepSeek V4 Flash (deepseek-flash) or DeepSeek V3.2 (deepseek-chat); they will answer cheaper and faster.

A written-out solution path is needed where you have to verify why the answer is what it is: in analytics, auditable calculations, and debugging complex prompts.