Devstral Small vs Gemini Embedding 2.
model vs model · prices as of
Devstral Small takes 3 of 5 rounds against Gemini Embedding 2. It is also the cheaper of the two to run, at $0.24 per million tokens against $0.60.
Mistral AIwins 3 roundsDevstral Small
$0.24/M tokens
cheapest we can explain, at NanoGPT · $0.0396 per 1,000 pages
128k context · tool calling · open weights
GoogleGemini Embedding 2
$0.60/M tokens
Google list price, nothing below it is accounted for · $0.12 per 1,000 pages
8k context · reads images
List price in (lower wins)
$0.10/M
$0.20/M
List price out (lower wins)
$0.30/M
$0/M
Cheapest we can explain (lower wins)
$0.24/M at NanoGPT
none below list
Standard price (lower wins)
$0.60/M
$0.60/M
Cost per 1,000 pages (lower wins)
$0.0396
$0.12
Context
128k tokens
8k tokens
Open weights
yes
no
Tool calling
yes
no
Structured output
not stated
not stated
Reads images
no
yes
Reasoning
no
no
Released
10 Jul 2025
22 Apr 2026
Providers selling it
9
4
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
not stated
no provider to check
Providers selling both · 1
LLMTR$0.60 / $12.60
Devstral Small then Gemini Embedding 2, per million tokens
Go further
Add a third or fourth model, or switch the weighting, in the tool.
What moved in the last seven days
- Gemini Embedding 2
Nous Portal$0.16 / $0 $0.20 / $025% dearer ·
Frequently asked
- Which is cheaper, Devstral Small or Gemini Embedding 2?
- Devstral Small, at $0.24 per million tokens against $0.60 for Gemini Embedding 2, three parts input to one part output, read on 16 Sept 2026.
- Where is each one cheapest?
- Devstral Small: $0.24 at NanoGPT. Gemini Embedding 2: nobody goes below the maker's own price with a reason we can point at, so $0.60 is the number to beat.
- Can I buy both from one provider?
- Yes. One provider sells both: LLMTR. The cheapest of them for the two together is LLMTR.
- Which holds more context?
- Devstral Small, at 128k tokens against 8k.
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.