Cohere
Cohere provides the Command R family of models, designed for enterprise use cases like retrieval-augmented generation (RAG), search, and multilingual tasks. All models support 128K+ context windows and tool calling via an OpenAI-compatible API.
Getting an API Key
- Visit dashboard.cohere.com/api-keys
- Sign in or create a Cohere account
- Generate a new API key
- Paste the key into AISCouncil under Settings > AI Model > Cohere
Cohere does not offer a free API tier for production use. All models are pay-as-you-go. A trial key with rate limits may be available for testing.
API keys are stored locally in your browser (localStorage) and are never included in shared bot URLs.
Supported Models
| Model | Context | Max Output | Input Price | Output Price | Capabilities |
|---|---|---|---|---|---|
| Command R+ | 128K | 4K | $2.50/MTok | $10.00/MTok | Tools, code, streaming |
| Command R | 128K | 4K | $0.15/MTok | $0.60/MTok | Tools, code, streaming |
| Command A | 256K | 8K | $2.50/MTok | $10.00/MTok | Tools, code, streaming |
| Command R7B | 128K | 4K | $0.04/MTok | $0.15/MTok | Tools, streaming |
Prices are per million tokens (MTok). See cohere.com/pricing for the latest rates.
- Command A is the latest generation with 256K context -- best for complex RAG and enterprise workflows.
- Command R+ is the previous flagship -- still strong for grounded generation and tool use.
- Command R offers a great balance of quality and cost at just $0.15/MTok input.
- Command R7B is the smallest and cheapest -- ideal for high-throughput tasks where cost matters most.
RAG and Enterprise Search
Cohere models are specifically optimized for retrieval-augmented generation. They handle grounded answers well when provided with context documents in the system prompt or conversation. The models also support strong multilingual capabilities across 100+ languages.
Configuration
Select Cohere as the provider when creating a bot profile. The app connects to api.cohere.com/compatibility/v1 using the OpenAI-compatible format with Bearer authentication.
Tips for Best Results
- Cohere excels at RAG workflows. If you are building a bot that answers questions from a knowledge base, Command R models are purpose-built for this.
- Use Command R for cost-effective multilingual tasks. At $0.15/MTok input, it handles 100+ languages well.
- Try Command A for long-document analysis. Its 256K context window can hold substantial reference material.