Very long inputs
Choose Command A
Command A reads up to 256k tokens per request, against 16k for babbage-002.
babbage-002 vs Command A: price, context and capabilities
Choose Command A for very long inputs and the latest release; on the other points the two are level or the data is missing.
In short
- A price comparison is not possible: one of them has no comparable published price.
- Command A reads more per request: 256k tokens against 16k.
- Command A is the newer model (March 2025, against August 2023).
- Only babbage-002 offers batch api.
- Only Command A offers tool calling.
Which should you choose?
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Latest release
Choose Command A
Command A is the newer release (March 2025, against August 2023 for babbage-002).
Specifications side by side
| babbage-002 | Command A | |
|---|---|---|
| Provider | OpenAI | Cohere |
| Category | Language models | Language models |
| Status | Deprecated | Generally available |
| Headline price | No data | No data |
| Input | No data | No data |
| Output | No data | No data |
| Cached input | No data | No data |
| Context window | 16k | 256k |
| Max output | No data | No data |
| Released | August 2023 | March 2025 |
| Knowledge cutoff | September 2021 | No data |
| Licence | Proprietary | Proprietary |
| Capabilities | Structured output, Batch API | Tool calling, Structured output |
| API model id | babbage-002 | command-a-03-2025 |
Benchmarks side by side
Leader in bold.
| Benchmark | babbage-002 | Command A |
|---|---|---|
| Terminal-Bench Hard | No data | 0.8% |
| GPQA Diamond | No data | 52.7% |
| Humanity's Last Exam | No data | 4% |
| IFBench (instruction following) | No data | 36.5% |
| τ²-Bench (agentic tool use) | No data | 15.2% |
| Long-context reasoning (AA-LCR) | No data | 21.3% |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Frequently asked questions
Should I choose babbage-002 or Command A?
Choose Command A for very long inputs and the latest release; on the other points the two are level or the data is missing.
Which is cheaper, babbage-002 or Command A?
A price comparison is not possible: one of them has no comparable published price.
Which has the larger context window, babbage-002 or Command A?
Command A reads more per request: 256k tokens against 16k.
What is the difference between babbage-002 and Command A?
babbage-002 is made by OpenAI and Command A by Cohere. A price comparison is not possible: one of them has no comparable published price. Command A reads more per request: 256k tokens against 16k. Command A is the newer model (March 2025, against August 2023). Only babbage-002 offers batch api. Only Command A offers tool calling.
Which is newer, babbage-002 or Command A?
Command A was released on March 2025 and babbage-002 on August 2023.