OpenSearch LLM-Based Conversational Search Edition provides built-in high-performance large language models (LLMs) and allows you to add external LLMs from Alibaba Cloud Model Studio. You can flexibly select and integrate LLMs based on your business requirements to improve the diversity and adaptability of your system. This topic describes how to view the LLMs supported by OpenSearch LLM-Based Conversational Search Edition.
View LLMs
On the details page of an OpenSearch LLM-Based Conversational Search Edition instance, click Model Management in the left-side pane. On the Model Management page, click the LLM tab. On the LLM tab, view the information about LLMs, including the model name, model type, and creation time. The model types include built-in LLM, external LLM, and custom LLM. You can also perform operations such as Q&A tests on the LLMs.
Model name | Model type | Description |
qwen3-235b-a22b | Built-in LLM | This model is a new-generation Qwen series LLM that is extensively trained. Qwen3 has made significant breakthroughs in inference, instruction following, agent capability, and multi-language support, can support more than 100 languages and dialects, and has powerful multi-language understanding, inference, and generation capabilities. |
qwen-vl-max | Built-in LLM | This model is the most capable LVLM in the Qwen family. It provides multimodal RAG capabilities for images and videos, and offers the best performance on more complex tasks. |
qwen-vl-plus | Built-in LLM | This model is the most capable LVLM in the Qwen family. It provides multimodal RAG capabilities for images and videos, and offers the best response speed. |
opensearch-qwen | Built-in LLM | This model is tuned by using the search enhancement training data based on Qwen-14b. |
opensearch-qwen-plus | Built-in LLM | This model is tuned by using the search enhancement training data based on Qwen-72b and provides more stable performance on the same question. |
deepseek-r1 | Built-in LLM | This model provides better effects in scenarios such as mathematical inference. |
deepseek-r1-distill-qwen-7b | Built-in LLM | This model is from Alibaba Cloud Model Studio. |
deepseek-r1-distill-qwen-14b | Built-in LLM | This model is from Alibaba Cloud Model Studio. |
opensearch-llama2-13b | Built-in LLM | This model is tuned by using the search enhancement training data based on llama2-13b. |
opensearch-llama2-13b-turbo | Built-in LLM | This model is an upgraded model that is tuned by using the search enhancement training data based on llama2-13b. |
opensearch-llama2-70b | Built-in LLM | This model is tuned by using the search enhancement training data based on llama2-70b. |
opensearch-llama3-8b | Built-in LLM | This model is tuned by using the search enhancement training data based on llama3-8b. |
qwen-turbo | Built-in LLM | Qwen series model that provides the optimal speed and low costs. The model is suitable for simple tasks. |
qwen-plus | Built-in LLM | Qwen series model with balanced capabilities. The inference effect and speed of this model are better than those of Qwen-Max but poorer than those of Qwen-Turbo. The model is suitable for medium-complex tasks. |
qwen-max | Built-in LLM | Qwen series model that provides the best effect. The model is suitable for complex and multi-step tasks. |
qwen2-72b-instruct | Built-in LLM | Qwen series model with balanced capabilities. The inference effect and speed of this model are better than those of Qwen-Max but poorer than those of Qwen-Turbo. The model is suitable for medium-complex tasks. The model is of a stable version and will not be updated in the future. |
You can perform only Q&A tests on built-in LLMs.
You can view details of, perform Q&A tests on, modify, and delete external LLMs.
You can view details of, perform Q&A tests on, train, and delete custom LLMs. Only Professional Edition instances of OpenSearch LLM-Based Conversational Search Edition support custom LLMs.