Model Positioning and Capabilities
A role-playing text generation model in the Qwen series, optimized for anthropomorphic interactions in Japanese. The official description states that it can follow character persona instructions, maintain character consistency, advance topics using context, listen and empathize, and recreate personalized characters. It is also optimized for understanding and expressing Japanese-localized knowledge, character anthropomorphism, and plot development. (docs.modelstudio.console.alibabacloud.com)
Use Cases
Suitable for Japanese-language virtual social interactions, game NPCs, interactions with virtual characters or IP characters, and anthropomorphic conversations in toys, in-vehicle devices, and similar settings. (alibabacloud.com)
Limitations
- Both input and output are text; the maximum context length is 8,192 tokens, the maximum input is 8,192 tokens, and the maximum output is 4,096 tokens. (help.aliyun.cn)
- The official model information page states that function calling, structured output, web search, prefix continuation, batch inference, and model tuning are not supported. (help.aliyun.cn)
- Official sources are inconsistent regarding cache support: the capability table on the model information page states that context caching is not supported, but the pricing table lists a cache-hit price, and the role-play documentation also mentions session caching. Actual availability should be determined by the console and invocation results. (help.aliyun.cn)
Pricing
The following are the original model invocation prices listed by the official sources, all per million tokens. Prices for different sites are listed independently and do not imply that they can be directly converted using an exchange rate. Query date: 2026-11-18.
- Alibaba Cloud China site, Chinese yuan (CNY): Singapore region, international deployment scope; input 3.67 yuan, output 10.275 yuan, cache-hit input 0.734 yuan. Official model information and pricing. (help.aliyun.com)
- Alibaba Cloud international site, US dollars (USD): Singapore region, international deployment scope; input 0.5 dollars, output 1.4 dollars, implicit cache-hit input 0.1 dollars. Official model information and pricing. (alibabacloud.com)
The actual applicability of cache pricing is affected by the documentation inconsistencies noted above; discounts and actual charges are subject to the corresponding platform console.
Release Date and Official Sources
- Release date: 2025-07-21, according to the date listed in the official model availability announcement. (docs.modelstudio.console.alibabacloud.com)
- Official model availability and update announcements
- Official role-play model documentation