Magistral Small vs Qwen2.5 Coder 32B Instruct.
model vs model · prices as of
Magistral Small takes 3 of 5 rounds against Qwen2.5 Coder 32B Instruct. It is also the cheaper of the two to run, at $1.05 per million tokens against $2.98.
Mistral AIwins 3 roundsMagistral Small
$1.05/M tokens
cheapest we can explain, at Eden AI · $0.126 per 1,000 pages
128k context · tool calling · open weights
Alibaba CloudQwen2.5 Coder 32B Instruct
$2.98/M tokens
OpenRouter list price, nothing below it is accounted for · $0.456 per 1,000 pages
33k context · open weights
List price in (lower wins)
$0.50/M
$0.66/M
List price out (lower wins)
$1.50/M
$1/M
Cheapest we can explain (lower wins)
$1.05/M at Eden AI
none below list
Standard price (lower wins)
$3/M
$2.98/M
Cost per 1,000 pages (lower wins)
$0.126
$0.456
Context
128k tokens
33k tokens
Open weights
yes
yes
Tool calling
yes
no
Structured output
not stated
no
Reads images
no
no
Reasoning
yes
no
Released
17 Mar 2025
11 Nov 2024
Providers selling it
6
4
Price we measure against
the maker's own
OpenRouter
Trains on your prompts (no is better)
not stated
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.
Frequently asked
- Which is cheaper, Magistral Small or Qwen2.5 Coder 32B Instruct?
- Magistral Small, at $1.05 per million tokens against $2.98 for Qwen2.5 Coder 32B Instruct, three parts input to one part output, read on 17 Sept 2026.
- Where is each one cheapest?
- Magistral Small: $1.05 at Eden AI. Qwen2.5 Coder 32B Instruct: nobody goes below OpenRouter's own price with a reason we can point at, so $2.98 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?
- Magistral Small, at 128k tokens against 33k.
Prices exclude tax and are per million tokens, read 17 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.