Fugu Max vs Mixtral 8x22B Instruct
Prices checked
Fugu Max and Mixtral 8x22B Instruct split the 5 rounds this table counts: the maker's two prices, the cheapest price, the maker's blended price and context. The weighting you pick decides the rest.
Sakana AIFugu Max
1M context · tool calling · reads images
$3/M tokensSakana AI's own price · $1.56 per 1,000 pages
Mistral AIMixtral 8x22B Instruct
66k context · tool calling · open weights
$0.60/M tokensCheapest price, at Nscale · $0.40 per 1,000 pages
Maker's price in (lower wins)
$2/M
$2/M
Maker's price out (lower wins)
$6/M
$6/M
Cheapest price (lower wins)
none below the maker
$0.60/M at Nscale
Maker's price, blended (lower wins)
$3/M
$3/M
Cost per 1,000 pages (lower wins)
$1.56
$0.40
Context
1M tokens
66k tokens
Open weights
not published
yes
Tool calling
yes
yes
Structured output
yes
yes
Reads images
yes
no
Reasoning
yes
no
Released
11 Sept 2026
17 Apr 2024
Providers selling it
11
9
Price we measure against
the maker's own
the maker's own
Trains on your prompts (no is better)
no provider to check
not published
Providers selling both · 5
Kilo Gateway$3 / $3
NanoGPT$3 / $3
Nous Portal$3 / $3
OpenRouter$3 / $3- AI/ML API$3.17 / $3.17
Fugu Max, then Mixtral 8x22B Instruct, per million tokens.
Go further
Add a third or fourth model, or change the weighting, in the tool.
Frequently asked
- Which is cheaper, Fugu Max or Mixtral 8x22B Instruct?
- Mixtral 8x22B Instruct, at $0.60 per million tokens against $3 for Fugu Max, blended 3 in : 1 out, read on 31 Oct 2026.
- Where is each one cheapest?
- Fugu Max: nobody sells it below the maker's own price with a stated reason, so $3 is the number to beat. Mixtral 8x22B Instruct: $0.60 at Nscale.
- Can I buy both from one provider?
- Yes. 5 providers sell both: Kilo Gateway, NanoGPT, Nous Portal, OpenRouter, AI/ML API. The cheapest of them for the two together is Kilo Gateway.
- Which holds more context?
- Fugu Max, at 1M tokens against 66k.
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