Model Catalog
The Deka LLM service in the Service Portal offers 11 models. The following table lists the available Deka LLM models that can be used.
Embeddings & Search
Models designed for semantic search, vector embeddings, and Retrieval Augmented Generation (RAG) use cases.
BAAI
bge-multilingual-gemma2
Vector embedding model for semantic search, similarity tasks, and text reranking (multilingual support).
Alibaba
qwen3-embedding-4b
Model designed to generate high-quality vector representations of text for semantic search, similarity matching, clustering, and Retrieval-Augmented Generation (RAG) applications.
Text Generation
Models designed for chatbot and virtual assistant development, summarization, documentation generation, and general natural language processing (NLP) use cases.
gemma-3-27b-it
Large language model for text generation, completion, and various NLP tasks (7B parameters).
Alibaba
qwen25-72b-instruct
Conversational AI model designed for interactive dialogue and instruction following (instruction-tuned).
Alibaba
qwen3-30b-a3b-instruct-2507
Conversational AI model designed for interactive dialogue and instruction following (instruction-tuned).
OpenAI OSS
gpt-oss-20b
Conversational AI model designed for interactive dialogue and instruction following (20B parameters).
GOTO
sahabat-ai-v2-70b-it
Large language model for text generation, completion, and various NLP tasks.
NVIDIA
nemotron-3-nano-30b-a3b
Optimized for text generation, conversational AI, and automation, offering efficient performance, low latency, and suitability for large-scale enterprise deployment.
OpenAI
whisper-large-v3
Large language model for text generation, completion, and various NLP tasks
Vision & Multimodal
Models designed to process and understand multimodal inputs, including images and text.
Alibaba
qwen25-vl-7b-instruct
Conversational AI model designed for interactive dialogue and instruction following (instruction-tuned)
Meta
llama-4-maverick-instruct
Vision-language model capable of understanding and analyzing images with text (instruction-tuned)
Last updated