82 lines
3.3 KiB
Markdown
82 lines
3.3 KiB
Markdown
# Available watsonx.ai Models
|
|
|
|
This document lists all foundation models currently available in watsonx.ai.
|
|
|
|
## Chat & Text Generation Models
|
|
|
|
| Model Label | Model ID | Provider | Parameters | Max Context |
|
|
|------------|----------|----------|------------|-------------|
|
|
| granite-3-8b-instruct | ibm/granite-3-8b-instruct | IBM | 8B | 8,192 |
|
|
| granite-3.1-8b-instruct | ibm/granite-3.1-8b-instruct | IBM | 8B | 131,072 |
|
|
| granite-4-h-small | ibm/granite-4-h-small | IBM | - | 131,072 |
|
|
| llama-3-3-70b-instruct | meta-llama/llama-3-3-70b-instruct | Meta | 70B | 131,072 |
|
|
| llama-guard-3-11b-vision | meta-llama/llama-guard-3-11b-vision | Meta | 11B | 131,072 |
|
|
| mistral-large-2512 | mistral-large-2512 | Mistral AI | 675B (41B active) | - |
|
|
| mistral-medium-2505 | mistralai/mistral-medium-2505 | Mistral AI | - | 131,072 |
|
|
| mistral-small-3-1-24b-instruct-2503 | mistralai/mistral-small-3-1-24b-instruct-2503 | Mistral AI | 24B | 128,000 |
|
|
| gpt-oss-120b | openai/gpt-oss-120b | OpenAI | 120B (5.1B active) | 131,072 |
|
|
|
|
## Embedding Models
|
|
|
|
| Model Label | Model ID | Provider | Dimensions | Max Tokens |
|
|
|------------|----------|----------|------------|------------|
|
|
| all-minilm6-v2 | sentence-transformers/all-minilm-l6-v2 | sentence-transformers | 384 | 512 |
|
|
| slate-125m-english-rtrvr | ibm/slate-125m-english-rtrvr | IBM | - | - |
|
|
| slate-30m-english-rtrvr | ibm/slate-30m-english-rtrvr | IBM | - | - |
|
|
|
|
## Model Capabilities
|
|
|
|
### IBM Granite Models
|
|
- **granite-3-8b-instruct**: General-purpose instruction-following model
|
|
- **granite-3.1-8b-instruct**: Enhanced version with extended context (131K tokens)
|
|
- **granite-4-h-small**: Latest Granite 4 model with advanced capabilities
|
|
|
|
### Meta Llama Models
|
|
- **llama-3-3-70b-instruct**: Large instruction-tuned model with 70B parameters
|
|
- **llama-guard-3-11b-vision**: Multimodal model with vision capabilities
|
|
|
|
### Mistral Models
|
|
- **mistral-large-2512**: State-of-the-art MoE model with 675B total parameters
|
|
- **mistral-medium-2505**: Multimodal model with vision and extended context
|
|
- **mistral-small-3-1-24b-instruct-2503**: Efficient 24B parameter model with vision
|
|
|
|
### OpenAI Models
|
|
- **gpt-oss-120b**: Open-weight model optimized for reasoning and agentic tasks
|
|
|
|
## Supported Tasks
|
|
|
|
All chat models support:
|
|
- Question Answering
|
|
- Summarization
|
|
- Retrieval Augmented Generation (RAG)
|
|
- Classification
|
|
- Text Generation
|
|
- Code Generation
|
|
- Extraction
|
|
- Translation
|
|
- Function Calling
|
|
|
|
## Usage Notes
|
|
|
|
1. **Context Length**: Models with 131K+ context length support long-document processing
|
|
2. **Vision Models**: Models with `image_chat` function support multimodal inputs
|
|
3. **Function Calling**: Most models support OpenAI-compatible function calling
|
|
4. **Multilingual**: Many models support multiple languages (see model specifications)
|
|
|
|
## Model Selection Guide
|
|
|
|
- **For general chat**: `ibm/granite-3-8b-instruct` or `meta-llama/llama-3-3-70b-instruct`
|
|
- **For long context**: `ibm/granite-3.1-8b-instruct` or `mistralai/mistral-medium-2505`
|
|
- **For vision tasks**: `meta-llama/llama-guard-3-11b-vision` or `mistralai/mistral-medium-2505`
|
|
- **For embeddings**: `sentence-transformers/all-minilm-l6-v2` or `ibm/slate-125m-english-rtrvr`
|
|
- **For reasoning**: `openai/gpt-oss-120b` or `mistral-large-2512`
|
|
|
|
## API Version
|
|
|
|
All models use API version: `2024-02-13`
|
|
|
|
---
|
|
|
|
*Last updated: 2026-02-23*
|
|
*Source: watsonx.ai Foundation Model Specs API*
|