|
Service Category
|
Service Description
|
|
Text Embedding
|
-
OpenSearch Text Embedding Service-001: Supports over 40 languages. Maximum input text length is 300 tokens. Output vector dimensions are 1536.
ID: ops-text-embedding-001
-
OpenSearch General Text Embedding Service-002: Supports over 100 languages. Maximum input text length is 8192 tokens. Output vector dimensions are 1024.
ID: ops-text-embedding-002
-
OpenSearch Text Embedding Service-Chinese-001: Supports Chinese only. Maximum input text length is 1024 tokens. Output vector dimensions are 768.
ID: ops-text-embedding-zh-001
-
OpenSearch Text Embedding Service-English-001: Supports English only. Maximum input text length is 512 tokens. Output vector dimensions are 768.
ID: ops-text-embedding-en-001
|
|
Content Generation
|
-
OpenSearch-Qwen-Turbo: Built on the qwen-turbo large language model. Uses supervised fine-tuning and retrieval-augmented generation to improve safety and relevance.
ID: ops-qwen-turbo
-
Qwen-Turbo: The fastest and most cost-effective model in the Qwen series. Best for simple jobs. For more information, see Model List.
ID: qwen-turbo
-
Qwen-Plus: Balanced performance. Offers inference quality, cost, and speed between Qwen-Max and Qwen-Turbo. Best for medium-complexity tasks. For more information, see Model List.
ID: qwen-plus
-
Qwen-Max: The highest-performing model in the Qwen series. Best for complex, multi-step tasks. For more information, see Model List.
ID: qwen-max
-
DeepSeek-R1: A large language model optimized for complex reasoning tasks. Excels at understanding complex instructions and delivering accurate results.
ID: deepseek-r1
-
DeepSeek-V3: A Mixture of Experts (MoE) model. Strong in long-context understanding, coding, math, encyclopedic knowledge, and Chinese language tasks.
ID: deepseek-v3
-
DeepSeek-R1-distill-qwen-7b: A Qwen-7B model fine-tuned using training samples generated by DeepSeek-R1. Built using knowledge distillation.
ID: deepseek-r1-distill-qwen-7b
-
DeepSeek-R1-distill-qwen-14b: A Qwen-14B model fine-tuned using training samples generated by DeepSeek-R1. Built using knowledge distillation.
ID: deepseek-r1-distill-qwen-14b
|