Command A vs GPT Audio.
model vs model · prices as of
Command A takes 1 of 5 rounds against GPT Audio. Both cost $17.50 per million tokens to run.
Coherewins 1 roundsCommand A
$17.50/M tokens
Cohere list price, nothing below it is accounted for · $2.10 per 1,000 pages
256k context · tool calling · open weights
OpenAIGPT Audio
$17.50/M tokens
OpenAI list price, nothing below it is accounted for · $2.10 per 1,000 pages
128k context · tool calling
List price in (lower wins)
$2.50/M
$2.50/M
List price out (lower wins)
$10/M
$10/M
Cheapest we can explain (lower wins)
none below list
none below list
Standard price (lower wins)
$17.50/M
$17.50/M
Cost per 1,000 pages (lower wins)
$2.10
$2.10
Context
256k tokens
128k tokens
Open weights
yes
not stated
Tool calling
yes
yes
Structured output
not stated
yes
Reads images
no
no
Reasoning
no
no
Released
13 Mar 2025
19 Jan 2026
Providers selling it
3
4
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
no provider to check
no provider to check
Providers selling both · 0
No provider we read sells both, so running the two means two accounts.
Go further
Add a third or fourth model, or switch the weighting, in the tool.
What moved in the last seven days
- GPT Audio
Nous Portal$2 / $8 $2.50 / $1025% dearer ·
Frequently asked
- Which is cheaper, Command A or GPT Audio?
- Neither: both run at $17.50 per million tokens, three parts input to one part output, read on 16 Sept 2026.
- Where is each one cheapest?
- Command A: nobody goes below the maker's own price with a reason we can point at, so $17.50 is the number to beat. GPT Audio: nobody goes below the maker's own price with a reason we can point at, so $17.50 is the number to beat.
- Can I buy both from one provider?
- No provider we read sells both, so running the two means two accounts and two bills.
- Which holds more context?
- Command A, at 256k tokens against 128k.
Prices exclude tax and are per million tokens, read 16 Sept 2026 from each provider's own catalog. The standard price blends three parts input to one part output; a page is 600 tokens in and 60 out. Bars are scaled to the larger value in each row.