Skip to content

docs/command-a and docs/models disagree on command-a-plus-05-2026's context window, max output, and hardware #794

Description

@AndrewAvery7

docs.cohere.com/docs/command-a states, for model ID command-a-plus-05-2026:

Context Window: 256,000 tokens
Max Output Tokens: 8,000 tokens
...
Hardware: Requires two GPUs to run (A100s / H100s)

But docs.cohere.com/docs/models' own row for that same model ID states:

command-a-plus-05-2026 | Live | ...can fit on 1 x B200 or 2 x H100 GPUs... | Context 128k | Max output 64k

Different context window, different max output, different hardware — same model ID, two Cohere docs pages.

For what it's worth, the numbers on /docs/command-a (256k / 8k / two A100s-H100s) match /docs/models' separate row for command-a-03-2025 almost exactly. Looks like the /docs/command-a page body wasn't updated when its Model ID field was changed to the new Command A+ id — the surrounding prose reads like it's still describing the older model.

Secondary, lower-confidence note: docs.cohere.com/docs/how-does-cohere-pricing-work says per-token pricing for Command A+ lives on cohere.com/pricing, but that page currently has no row for command-a-plus-05-2026 or command-a-03-2025 at all — only a legacy-models FAQ answer covering older Command models. Flagging in case it's related, but the spec mismatch above is the main thing.

Happy to provide more detail if useful.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions