Model · 001
Qwen3.8-27B
A dense 27-billion-parameter vision-language model from Alibaba, released under Apache 2.0. Tannit serves it as tannit/qwen3.8-27b.
Specification
Datasheet
| Base model | Qwen3.8-27B |
|---|---|
| Developer | Alibaba |
| Licence | Apache License 2.0 |
| Architecture | Dense, 27B parameters |
| Context window | 262,144 tokens |
| Max output tokens | [[MAX_OUTPUT_TOKENS]] |
| Input | Text, image, video |
| Output | Text |
| Reasoning control | reasoning_effort |
| Tool calling | [[TOOL_CALLING]] |
| Structured output | [[STRUCTURED_OUTPUT]] |
| Streaming | Supported |
| Regions | EU, US, CA, KR, CN, IL |
| Price · input | [[PRICE_INPUT]] / 1M tokens |
| Price · output | [[PRICE_OUTPUT]] / 1M tokens |
Integration
Calling the endpoint
The API follows the OpenAI chat completions schema. Any client that speaks it will work once the base URL is changed.
# Python — openai SDK from openai import OpenAI client = OpenAI( base_url="[[API_BASE_URL]]", api_key="$TANNIT_API_KEY", ) resp = client.chat.completions.create( model="tannit/qwen3.8-27b", messages=[{"role": "user", "content": "Summarise this repo."}], stream=True, )
Discovery
A /models endpoint returns the served model list with pricing, context length, maximum output tokens, supported features and datacenter regions.
Attribution
Licence and provenance
Qwen3.8-27B was developed and released by Alibaba under the Apache License 2.0. Tannit AI serves the model without modification to its weights and is not affiliated with, sponsored by, or endorsed by Alibaba. Tannit's proprietary work is in the hardware and inference stack the model runs on, not in the model itself.