Cohere is an enterprise-focused AI provider whose sweet spot is retrieval: its Embed 5 and Rerank 5 models are widely used to power RAG and search, and its Command models handle generation. If you're building a search or knowledge-base product rather than a chatbot, Cohere is often the cheaper, more accurate backbone. Here's what it costs and how to get your key.
| Modello | Inserisci $/1 milione | Produzione $/1 milione | Meglio per |
|---|---|---|---|
| Command R+ | $2.50 | $10.00 | Flagship generation, RAG |
| Command R cheap | $0.15 | $0.60 | High-volume RAG & tools |
| Embed 5 Pro | $0.12 per 1M tokens | Vector embeddings (replaces Embed v3) | |
| Embed 5 Fast cheap | $0.08 per 1M tokens | Faster, lower-cost embeddings | |
| Rerank 5 Pro | $0.05 per 1M tokens | Reordering search results (replaces old Rerank) | |
| Rerank 5 Fast cheap | $0.02 per 1M tokens | Faster, lower-cost reranking | |
→ Estimate the generation side on the Calcolatore dei costi dell'API AI, or model a retrieval app with the Calcolatore dei costi RAG.
Yes — Cohere issues free trial API keys with rate limits, suitable for evaluation and prototyping across Command, Embed and Rerank. You upgrade to a production key with billing when you need higher throughput. For other genuinely free options, compare Gemelli E Mistral.
1. Iscriviti a dashboard.cohere.com.
2. Open the Chiavi API page — a free trial key is created for you automatically.
3. Crea un production key and add billing when you outgrow the trial limits.
4. Use the official Cohere SDKs, or call the REST endpoints directly.
Test it with a simple request:
For chat/generation, use the Calcolatore dei costi dell'API AI to compare Command against GPT-4o, Claude and Gemini. For a full retrieval system — embeddings, vector store, rerank and generation — the Calcolatore dei costi RAG breaks down where the money actually goes. Note Rerank 5 bills per token processed, not per query like the old Rerank — a reranking-heavy app's bill now scales with document length, not just request count.
For pure chat, DeepSeek E Gemini Flash undercut Command R+. But for reranking and embeddings specifically, Cohere is often the price-performance leader — compare its Embed 5/Rerank 5 against OpenAI embeddings before switching. To access Cohere plus many others through one key, see OpenRouter.
Retrieval. Its Embed and Rerank models are built for search and RAG, often delivering better relevance per dollar than general-purpose chat models.
Yes — free, rate-limited trial keys for evaluation and prototyping, with paid production keys for scale.
Sign up at dashboard.cohere.com; a free trial key is generated automatically. Create a production key and add billing for higher limits.
Command R/R+ handle chat and tool use well, but for open-ended consumer chat the big general models may be cheaper. Cohere shines in RAG and enterprise search.
Not affiliated with Cohere. Prices are reference estimates — always verify on the official pricing page.