Cohere’s Command A advertises a $10 per 1M output token rate that looks unbeatable next to GPT‑5‑nano’s $0.4 and Claude‑sonnet‑5’s $10, yet its $2.5 per 1M input price dwarfs OpenAI’s $0.05 for gpt‑5‑nano and Claude‑3‑haiku’s $0.25. The contrast flips the economics depending on whether your workload reads more than it writes.
### How the token bills compare
| Vendor | Model | Input | Cached input | Output |
|--------|-------|-------|--------------|--------|
| Cohere | command‑a | $2.5 per 1M tokens | – | $10 per 1M tokens |
| OpenAI | gpt‑5‑nano | $0.05 per 1M tokens | $0.005 per 1M tokens | $0.4 per 1M tokens |
| OpenAI | gpt‑4o‑mini‑2024‑07‑18 | $0.15 per 1M tokens | $0.075 per 1M tokens | $0.6 per 1M tokens |
| Anthropic | claude‑sonnet‑5 | $2 per 1M tokens | $0.2 per 1M tokens | $10 per 1M tokens |
Cohere’s flat‑rate structure means every token—prompt or completion—costs the same. For generation‑heavy use cases such as long‑form article drafting or multi‑step code synthesis, the $10 output fee competes well with Claude‑sonnet‑5 and stays far below OpenAI’s $0.6‑$1.2 range for comparable models. Conversely, any pipeline that ingests millions of prompt tokens quickly feels the sting of a $2.5 input charge.
OpenAI spreads the cost across three levers. The ultra‑low $0.05 input price for gpt‑5‑nano makes it the cheapest option for prompt‑intensive workloads, and the cached‑input discount ($0.005) halves that price when prompts are reused. Even its higher‑tier gpt‑4o‑mini still charges only $0.15 input and $0.075 cached input, keeping read‑heavy scenarios inexpensive while keeping output at $0.6.
Anthropic sits in the middle. Its $2 input and $0.2 cached input rates are higher than OpenAI but lower than Cohere, while the $10 output matches Cohere’s. This makes Anthropic a viable choice for teams that need a balanced spend between ingestion and generation without committing to the extremes of either vendor.
### Which buyer should lean where
Start‑ups or analytics platforms that flood a model with short logs, alerts, or query strings will see the greatest cost savings with OpenAI’s gpt‑5‑nano or gpt‑4o‑mini because the input bill is an order of magnitude lower than any competitor. Enterprises that monetize large‑scale content creation—marketing agencies, e‑learning publishers, or code‑generation services—will benefit from Cohere’s $10 output rate, which lets them price their output‑heavy offerings competitively while still undercutting Anthropic’s identical output cost. For teams that need a middle ground—moderate prompt volume paired with sizable outputs—Anthropic’s claude‑sonnet‑5 offers a predictable split without the high input surcharge of Cohere.
In short, choose OpenAI if your product lives on cheap prompt ingestion, pick Cohere if your revenue hinges on massive generation per request, and consider Anthropic when you want a balanced price sheet without extreme input or output spikes. Pricing verified on 2026-09-12.