GPT-5 Image Mini vs Qwen2.5 72B Instruct.
model vs model · prices as of
GPT-5 Image Mini takes 3 of 5 rounds against Qwen2.5 72B Instruct. Qwen2.5 72B Instruct counters on price: $1.09 per million tokens against $9.50, for 131k context instead of 400k.
GPT-5 Image Mini
$9.50/M tokens
OpenAI list price, nothing below it is accounted for · $1.62 per 1,000 pages
400k context · reads images
Qwen2.5 72B Instruct
$1.09/M tokens
cheapest we can explain, at Requesty · $0.162 per 1,000 pages
131k context · tool calling · open weights
List price in (lower wins)
$2.50/M
$1.40/M
List price out (lower wins)
$2/M
$5.60/M
Cheapest we can explain (lower wins)
none below list
$1.09/M at Requesty
Standard price (lower wins)
$9.50/M
$9.80/M
Cost per 1,000 pages (lower wins)
$1.62
$0.162
Context
400k tokens
131k tokens
Open weights
not stated
yes
Tool calling
no
yes
Structured output
yes
not stated
Reads images
yes
no
Reasoning
yes
no
Released
16 Oct 2025
1 Sept 2024
Providers selling it
3
15
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
no provider to check
not stated
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.
Frequently asked
- Which is cheaper, GPT-5 Image Mini or Qwen2.5 72B Instruct?
- Qwen2.5 72B Instruct, at $1.09 per million tokens against $9.50 for GPT-5 Image Mini, three parts input to one part output, read on 16 Sept 2026.
- Where is each one cheapest?
- GPT-5 Image Mini: nobody goes below the maker's own price with a reason we can point at, so $9.50 is the number to beat. Qwen2.5 72B Instruct: $1.09 at Requesty.
- 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?
- GPT-5 Image Mini, at 400k tokens against 131k.
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.