Qwen with a 262K window: from Qwen3 Max to Qwen3.8 27B

Qwen3 Max was the top Qwen with a 262K-token context. In the Yenisei catalog its place has been taken by Qwen3.8 27B (qwen3.8-27b), with the same window, reasoning, and images as input. It is a convenient middle ground: more than regular chat models offer, but without the weight and cost of a million-token window.

Specifications

• 262K-token context: large documents, long conversations, several files at once.
• Tools: full function calling for agents and integrations.
• Reasoning and images: Qwen3.8 27B has both.
• High-quality Russian: clean, without anglicisms.

Where 262K is enough

Most long-context tasks don't need a million tokens. Going through technical documentation, analyzing a quarterly report, working with several related code files: all of this fits into 262K with room to spare. And a window of this size is processed faster and cheaper than a million-token one.

curl https://api.yenisei.ru/v1/chat/completions \
  -H "Authorization: Bearer $YENISEI_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-27b",
    "messages": [
      {"role": "user", "content": "Here is the entire API documentation. Put together a short guide to authentication."}
    ]
  }'

How to choose between 262K and a million

A simple rule: if your material fits into 262K, use Qwen3.8 27B; it will be faster and cheaper. If you regularly hit the ceiling and really need to put in a whole book or an entire repository, then go for Qwen3.8 Flash (qwen-flash) or Qwen3.8 2.4T (qwen3.8-2.4t) with their million tokens. There is no point paying for a million-token window while using a tenth of it.