Qwen 3.8 27B

Qwen 3.8 27B lists base rates of $0.50 per million input tokens and $3 per million output tokens. Estimate both input and output token volume and any additional charges for your workload. Hosted qwen3.8-27b International lists $0.50/$3.00 per MTok; China lists $0.424/$1.696. QwenCloud implicit cache $0.10, explicit creation $0.625/read $0.05. Its hosted card lists 1M context/131K maximum response; the downloadable Apache-2.0 checkpoint has 262,144 native context. August 19 is the managed SKU listing, not proof of checkpoint release day.

Qwen 3.8 27B accepts about 262K tokens of context. That only matters if you routinely send very long documents, large codebases, or multi-turn histories that approach that limit.

Qwen 3.8 27B is a budget tier from Qwen. It suits self-hosting on a single node, multimodal input at small scale, long-context on modest hardware. Exact checkpoint release day and hard output cap remain unverified. Thinking defaults to xhigh and supports low/medium/xhigh or non-thinking; historical thinking is preserved by default. Runtime memory, quantization and context scaling need target-hardware verification. Managed 1M context/output limits do not transfer to native weights; video is input, text is output.

Qwen 3.8 27B publishes open weights, but self-hosting depends on the licence, hardware footprint, quantisation quality, and serving stack. A hosted API is often cheaper until you have measured throughput and concurrency on your own hardware.