How to Get Your Cohere API Key (Best-in-Class Embeddings)
The specialist for embeddings, reranking, and RAG pipelines

Cohere is the specialist for embeddings and reranking. If your use case involves RAG, semantic search, or any retrieval system, Cohere's embed models consistently top the leaderboards. The trial key is free forever for development - no credit card required.

Prerequisites

• A dashboard.cohere.com account (free to sign up)
• No credit card required - the trial key works indefinitely for low-volume testing

Get your key in 3 steps

1. Sign up and log in

Go to dashboard.cohere.com and create an account. Email signup is straightforward, no credit card needed.

2. Open "API Keys" in the left sidebar

Once logged in, find API Keys in the left sidebar navigation. Cohere creates a default trial key for your account automatically.

3. Copy your key

Your trial key is displayed in the API Keys section. Click to copy it. You can also create additional named keys for specific projects. The trial key is rate-limited but has no expiration - ideal for development and testing.

Models unlocked with your key

Cohere's strength is its retrieval stack. Here's what you get:

LLMs:

• Command R+ - Cohere's flagship model, optimized for RAG and tool use, strong on long-context tasks
• Command R - faster and cheaper, still RAG-optimized
• Command Light - lightweight for latency-sensitive applications

Embeddings - Cohere's core superpower:

• embed-v4.0 - latest generation, multilingual, 1024 dimensions, top performer on retrieval benchmarks
• embed-v3-multilingual - excellent cross-lingual retrieval across 100+ languages
• embed-v3-english - optimized for English-only retrieval tasks

Reranking - critical for RAG quality:

• rerank-v3.5 - takes a query and a list of documents, reorders them by actual relevance. This is the step most RAG pipelines skip, and it makes a measurable difference in answer quality.

Classification and clustering:

• Text classification and zero-shot classification
• Semantic clustering for document organization

Why embeddings matter - and why Cohere leads

Most RAG pipelines retrieve too many candidates and trust the initial vector similarity score too much. A two-step approach - embed first, rerank second - dramatically improves answer quality:

1. Vector search returns the top 50-100 candidates (fast, approximate)
2. rerank-v3.5 re-scores those candidates against the actual query (slower, precise)

The reranker understands context that embedding similarity misses. Cohere's rerank models are specialized for this exact task and consistently outperform generic alternatives.

For multilingual applications, embed-v4.0 produces robust vector representations across languages - a concrete advantage over models trained primarily on English data.

Pricing

Cohere's trial key is free with rate limits - no time limit, no credit card. For production:

• embed-v4.0: very competitive per-token pricing
• rerank-v3.5: priced per search unit
• Command R+: competitive with comparable frontier models

The trial key is generous enough for building and benchmarking your entire RAG pipeline before committing to production pricing.

Security tips

• Never hardcode your key in source code or commit it to a Git repository
• Use environment variables (.env files) in all your projects
• Create a separate production key when you move beyond development - trial and production keys should be distinct
• Monitor usage from the Cohere dashboard to catch unexpected spikes

Connect your key to HiWay2LLM

Once you have your Cohere key, bring it to HiWay2LLM in seconds. All your Cohere calls - embeddings, reranking, Command R - route through a unified endpoint with cost tracking and full observability built in.