LLMTR / llmtr/qwen3-8-27b

Qwen 3.8 27B - access through LLMTR

LLMTR Qwen 3.8 27B is a first-party language model hosted in Turkey; requests are never forwarded to a third-party provider. It is invoked through the OpenAI-compatible Chat Completions API and offers a 262,144-token context window (prompt and completion together, per request). It makes tool/function calls, and with `response_format` it returns a JSON object or output that conforms to the JSON schema you supply. Thinking is off by default: add `:think` to the model id or send `reasoning: true` in the body to turn it on. With thinking on, the reasoning is returned in `reasoning_content` and is spent from your `max_tokens` budget, so give `max_tokens` at least 4,096. A repeated prompt prefix can be served from cache; only the tokens reported as `cached_tokens` are billed at the discounted rate. It accepts text input only; image, audio and file input are not supported.