DeepSeek V3.1 Terminus vs GPT-5.4 nano
Prices checked
DeepSeek V3.1 Terminus takes 3 of 5 rounds against GPT-5.4 nano. It is also the cheaper of the two to run, at $0.36 per million tokens against $0.42.
DeepSeekWins 3 roundsDeepSeek V3.1 Terminus
164k context · tool calling · open weights
$0.36/M tokensCheapest price, at NanoGPT · $0.19 per 1,000 pages
OpenAIGPT-5.4 nano
400k context · tool calling · reads images
$0.42/M tokensCheapest price, at Poe · $0.18 per 1,000 pages
Maker's price in (lower wins)
$0.27/M
$0.20/M
Maker's price out (lower wins)
$1/M
$1.25/M
Cheapest price (lower wins)
$0.36/M at NanoGPT
$0.42/M at Poe
Maker's price, blended (lower wins)
$0.45/M
$0.46/M
Cost per 1,000 pages (lower wins)
$0.19
$0.18
Context
164k tokens
400k tokens
Open weights
yes
no
Tool calling
yes
yes
Structured output
yes
yes
Reads images
no
yes
Reasoning
yes
yes
Released
22 Sept 2025
17 Mar 2026
Providers selling it
15
35
Price we measure against
OpenRouter
the maker's own
Trains on your prompts (no is better)
not published
not published
Providers selling both · 7
NanoGPT$0.36 / $0.46
Eden AI$0.44 / $0.46
Abacus$0.45 / $0.46
Kilo Gateway$0.45 / $0.46
Nous Portal$0.45 / $0.46
OpenRouter$0.45 / $0.46
TrustedRouter$0.90 / $0.49
DeepSeek V3.1 Terminus, then GPT-5.4 nano, per million tokens.
Go further
Add a third or fourth model, or change the weighting, in the tool.
What moved this week
- TrustedRouter · 85% dearer · $0.60 / $1.81 $0.32 / $1
Frequently asked
- Which is cheaper, DeepSeek V3.1 Terminus or GPT-5.4 nano?
- DeepSeek V3.1 Terminus, at $0.36 per million tokens against $0.42 for GPT-5.4 nano, blended 3 in : 1 out, read on 31 Oct 2026.
- Where is each one cheapest?
- DeepSeek V3.1 Terminus: $0.36 at NanoGPT. GPT-5.4 nano: $0.42 at Poe.
- Can I buy both from one provider?
- Yes. 7 providers sell both: NanoGPT, Eden AI, Abacus, Kilo Gateway, Nous Portal, OpenRouter, and more. The cheapest of them for the two together is NanoGPT.
- Which holds more context?
- GPT-5.4 nano, at 400k 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.