command-a

By price, command-a ranks 55th cheapest of 68 on input and 51st on output. Its quality score of 83 ranks 44th of 68, and on value (quality score ÷ output price) it ranks 51st of 68. Its blended rate sits 49% above the catalog average.

In real billing terms: a support chatbot handling 50,000 turns a month (2,000 input / 500 output tokens each) costs about $500 per month on command-a, while a RAG workload of 20,000 queries (8,000 input / 800 output tokens each) runs about $560. Because output is priced 4.0× input here, long-generation tasks scale cost fastest — cap output length before you optimise anything else.

A 256K context window holds roughly 341 pages of A4 text (at ~750 tokens per page), ranking 37th of 68 here. That covers typical RAG and multi-turn chat, though book-length inputs still need retrieval and chunking. Output runs at about 70 tok/s (50th fastest of 68), acceptable interactively, though long responses will feel slow.

We track 2 Cohere models, and command-a's blended rate sits 0% above that provider's average. If your task tolerates a little less headroom, qwen3.5-flash (Alibaba) blends about 99% cheaper at a quality score of 82, and is worth benchmarking first. Going the other way, Hunyuan HY3 scores 87 on quality for about 98% less.

The closest model in Cohere's line-up is command-r ($2.50 input / $10.00 output, quality 75). Against it, command-a scores 8 points higher on quality, runs 50 tok/s slower (about 42%), has a 128K larger context window. When two models sit this close on price, the deciding factor is usually the quality ceiling — run both against your real prompts before picking.