DeepSeek V3.2 vs Mercury 2.5
Prices checked
Mercury 2.5 takes 3 of 5 rounds against DeepSeek V3.2. It is also the cheaper of the two to run, at $0.07 per million tokens against $0.23.
DeepSeekDeepSeek V3.2
164k context · tool calling · open weights
$0.23/M tokensCheapest price, at NanoGPT · $0.14 per 1,000 pages
InceptionWins 3 roundsMercury 2.5
260k context · tool calling
$0.07/M tokensCheapest price, at Nous Portal · 3.3¢ per 1,000 pages
Maker's price in (lower wins)
$0.28/M
$0.20/M
Maker's price out (lower wins)
$0.40/M
$0.75/M
Cheapest price (lower wins)
$0.23/M at NanoGPT
$0.07/M at Nous Portal
Maker's price, blended (lower wins)
$0.31/M
$0.34/M
Cost per 1,000 pages (lower wins)
$0.14
3.3¢
Context
164k tokens
260k tokens
Open weights
yes
no
Tool calling
yes
yes
Structured output
yes
yes
Reads images
no
no
Reasoning
yes
yes
Released
1 Dec 2025
8 Sept 2026
Providers selling it
49
10
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
not published
not published
Providers selling both · 3
Nous Portal$0.23 / $0.07
Kilo Gateway$0.29 / $0.34
Venice$0.37 / $0.42
DeepSeek V3.2, then Mercury 2.5, per million tokens.
Go further
Add a third or fourth model, or change the weighting, in the tool.
What moved this week
- Kilo Gateway · 8% cheaper · $0.26 / $0.38 $0.28 / $0.42
OpenRouter · 8% cheaper · $0.26 / $0.38 $0.28 / $0.42
io.net Intelligence · 8% cheaper · $0.95 / $1.43 $1 / $1.65
Eden AI · 114% dearer · $0.50 / $1.50 $0.30 / $0.50
Frequently asked
- Which is cheaper, DeepSeek V3.2 or Mercury 2.5?
- Mercury 2.5, at $0.07 per million tokens against $0.23 for DeepSeek V3.2, blended 3 in : 1 out, read on 31 Oct 2026.
- Where is each one cheapest?
- DeepSeek V3.2: $0.23 at NanoGPT. Mercury 2.5: $0.07 at Nous Portal.
- Can I buy both from one provider?
- Yes. 3 providers sell both: Nous Portal, Kilo Gateway, Venice. The cheapest of them for the two together is Nous Portal.
- Which holds more context?
- Mercury 2.5, at 260k tokens against 164k.
Prices exclude tax, read 31 Oct 2026 from each provider's own catalog. They are per M tokens, blended 3 in : 1 out, and a page is 600 tokens in and 60 out. Bars are scaled to the larger value in each row.
Related comparisons