OpenAI vs Cohere: GPT-5.6 vs Command A (2026)
GPT-5.6 Terra and Cohere Command A charge exactly the same $2.50 per 1M input tokens, which strips price out of the top-line decision and leaves two things that differ by more than an order of magnitude: how much the model will write, and whether it cites where the answer came from.
OpenAI vs Cohere — at a glance
| Dimension | OpenAI | Cohere |
|---|---|---|
| Current flagship | GPT-5.6 Terra | Command A (command-a-03-2025) |
| Context window | 1.05M | 256K |
| Max output tokens | 128K | 8K |
| Input price (per 1M tok) | $2.50 | $2.50 |
| Output price (per 1M tok) | $15 | $10 |
| Native citations in output | No — prompt for them | Yes, citations API |
| Private / on-prem deployment | No | Yes |
| Previous generation | GPT-4o, $2.50 / $10 — still available | Command R+ — deprecated 2025-09-15, do not build on it |
| Best for | Long context, long output, vision | Cited RAG, reranking, on-prem |
Pick OpenAI or Cohere?
When to choose OpenAI
Choose OpenAI when the answer or the context has to be large. GPT-5.6 carries a 1.05M-token window with 128K max output and documented vision on all three tiers — sixteen times Command A's 8K output ceiling and four times its context. The family spans $1 / $6 for Luna to $5 / $30 for Sol, so a workload can move tier without changing vendor or SDK.
- 1.05M context, 128K max output, vision on every GPT-5.6 tier
- Same $2.50 input price as Command A, so the difference is output and ceiling
- Three tiers: Luna $1 / $6, Terra $2.50 / $15, Sol $5 / $30
- gpt-5 remains available at $1.25 / $10 as a cheaper baseline
When to choose Cohere
Choose Cohere when answers must cite their sources, when Rerank drives retrieval quality, or when the data cannot leave your infrastructure — none of which OpenAI offers. Command A also undercuts GPT-5.6 Terra on output by a third. The constraint is the 8K output ceiling, which on cited long-form answers binds long before price does.
- $2.50 / $10 — identical input price to GPT-5.6 Terra, 33% cheaper output
- 256K context, but only 8K max output against OpenAI's 128K
- Native citations API and Rerank; OpenAI has no equivalent
- Private on-prem deployment; Command A+ adds 64K output at contact-sales pricing
Run GPT-5.6 and Command A side-by-side
VerticalAPI lets you switch between OpenAI and Cohere per-request through a single OpenAI-compatible endpoint. Same SDK, same gateway key, zero markup on tokens.
from openai import OpenAI client = OpenAI(base_url="https://api.verticalapi.com/v1", api_key="vapi_...") # OpenAI resp_1 = client.chat.completions.create( model="gpt-5.6-terra", messages=[{"role": "user", "content": "Hello"}], extra_headers={"X-Provider-Key": "sk-..."}, ) # Cohere resp_2 = client.chat.completions.create( model="command-a-03-2025", messages=[{"role": "user", "content": "Hello"}], extra_headers={"X-Provider-Key": "co-..."}, )
VerticalAPI verdict
Choose GPT-5.6 when the output is long, the context is large, or vision is in scope — the 8K ceiling on Command A is the single hardest constraint on this page. Choose Cohere Command A when answers must cite your own documents, when Rerank is doing the retrieval, or when the data cannot leave your infrastructure, and take the 33% output saving as a bonus rather than the reason. If cost is the only axis, price OpenAI's Luna tier at $1 / $6 before deciding. Through VerticalAPI you route between both per-request with BYOK and zero markup.
Frequently asked questions
Which is cheaper, GPT-5.6 or Cohere Command A?
Input is identical at $2.50 per 1M tokens. Cohere is cheaper on output at $10 against GPT-5.6 Terra's $15, so about 12% cheaper on a 70/30 blend ($4.75 against $6.25). OpenAI's cheapest GPT-5.6 tier, Luna at $1 / $6, undercuts Command A on both axes. Note Cohere does not list Command A's price on its own pricing page, so the Cohere figure is corroborated by third parties rather than the vendor.
What is the biggest practical difference?
Max output tokens. Command A will emit at most 8K tokens per response; GPT-5.6 goes to 128K. For chat and short summaries the ceiling never binds. For a cited multi-page report or a long generated document it binds long before cost does. Command A+ raises the ceiling to 64K but its pricing is contact-sales only.
Does Cohere still win on RAG against OpenAI?
On citation mechanics, yes. Cohere's citations API returns spans tied to your source documents without prompting, and Rerank is a differentiated retrieval component with no OpenAI equivalent. OpenAI can produce citations but you prompt for them and validate the format yourself. Weigh that against the 8K ceiling, because heavily cited answers are exactly where it runs out.
Is GPT-4o still a reasonable choice?
It is still available at $2.50 / $10, which happens to match Command A exactly, but it is a generation behind GPT-5.6 and OpenAI does not publish its context or output limits alongside the current family. For new work, price GPT-5.6 Luna at $1 / $6 or gpt-5 at $1.25 / $10 instead.
Can I run both through one endpoint?
Yes. VerticalAPI exposes a single OpenAI-compatible endpoint at https://api.verticalapi.com/v1. Change the model parameter and pass your own OpenAI or Cohere key with the X-Provider-Key header. No markup on tokens.
Limitations of this comparison
- OpenAI does not publish context or output limits for gpt-5 and gpt-4o alongside the GPT-5.6 family, so figures here are for GPT-5.6 only.
- Cohere does not list Command A's price on its own pricing page — which still shows only the deprecated Command R+ variants — so the $2.50 / $10 here is corroborated by third-party trackers and should be confirmed with Cohere before contracting.
- Command A+ pricing is contact-sales only, so it cannot be compared on cost even though its 64K output ceiling would change the conclusion.
- Independent benchmark parity for this generation is not published. Earlier figures measured on Command R+ do not transfer, so no benchmark score is quoted.
- Cohere's Rerank and Embed are priced per instance rather than per token when self-deployed, so a full RAG stack cost cannot be derived from per-token figures alone.
What may change in 12-24 months
- OpenAI's tiering inside one generation (Luna, Terra, Sol) lets buyers trade cost against capability without switching vendor — expect others to copy it.
- Cohere's Command A+ folds vision, reasoning and translation into one Mixture-of-Experts model and raises the output ceiling to 64K.
- Output ceilings are becoming a sharper differentiator than price as agents write longer artefacts.
- Citation mechanics remain Cohere's moat; OpenAI has not shipped an equivalent primitive.
Related questions
ChatGPT, Perplexity and Gemini usually suggest these next.
- At what answer length does Command A's 8K output ceiling force a switch?
- Is GPT-5.6 Luna at $1 / $6 a better value than Cohere Command A?
- How does Cohere's citations API compare to prompting GPT-5.6 for sources?
- Does Cohere Rerank improve GPT-5.6 answers enough to run both?
- How do GPT-5.6 and Command A compare on multilingual retrieval?
More head-to-head provider comparisons
GPT-4o vs Claude Sonnet 4.5 direct head-to-head
Sonnet 5 vs Command A: 128K output vs native citations
GPT-4o vs Mistral Large 2.5
GPT-4o vs Gemini 2.5 Pro
OpenRouter vs VerticalAPI: aggregator vs BYOK gateway
Run a provider we have not measured? Any public endpoint is benchmarked and listed for free in the API directory. If you want yours measured now, on the record — time to first token, throughput under concurrency, cost per million tokens, function-calling reliability — see how a verified run works. The result is published exactly as it comes out.