Models
Available base models and their capabilities
Text Models
| Model | Size | Description | Best For |
|---|---|---|---|
| mistral-7b-instruct | 7B | Instruction-tuned open-source model | Chatbots, Q&A, summarization |
| llama-3-8b | 8B | Meta's LLaMA 3 model | General-purpose text generation |
| llama-3-70b | 70B | LLaMA 3 large model with improved reasoning | Complex reasoning, long-form text |
Embedding Models
| Model | Size | Description | Best For |
|---|---|---|---|
| bge-large-en-v1.5 | 335M | High-quality embeddings for English text | Semantic search, RAG pipelines |
| gte-base | 110M | Lightweight embedding model | Client-side or low-latency apps |
Audio Models
| Model | Size | Description | Best For |
|---|---|---|---|
| whisper-small | 244M | Fast speech-to-text model | Real-time transcription |
| whisper-large-v3 | 1550M | More accurate Whisper model | High-quality transcription |
Model Naming Convention
Models follow the format: provider-model-size-version (e.g., dendrer-chat-13b-v1, meta-llama3-8b, mistral-7b-instruct-v0.2). Fine-tuned models use the prefix 'ft:' (e.g., ft:dendrer-chat-13b:custom-support-v1).