Models / Sakana AI / Fugu Ultra

Sakana AI · fugu

Fugu Ultra

Quality-first multi-agent model for hard research, analysis, and competitions

$5.00 in · $30.00 out

per 1M tokens · first-party price. Reference price from Sakana AI, the lab's own listing.

Provider input prices run from $5.00 to $7.50 per 1M tokens. Output prices run from $30.00 to $45.00.

Context
1,000,000
Max output
—
Cache read
$0.50
Cache write
—
Reasoning
Yes
Tool calls
Yes
Structured output
Yes
Vision
Yes
Open weights
No
License
—
Released
Jun 15, 2026
Updated in source
Jun 15, 2026
Knowledge
—
Providers
12

Higher rates apply above 200,000, 272,000 input tokens. The calculator switches to that rate when the input is longer.

Official model catalog

What this model costs per month

Prices are USD per 1M tokens. A month is 30 days. Cache hits use the cache-read rate when the model has one; otherwise the cache field is ignored.

monthly = requests/day × 30 × (input tokens × (1 − cache hit) × input price + input tokens × cache hit × cache-read price + output tokens × output price) ÷ 1,000,000

Providers

Base rates from Models.dev. Where a provider lists the model more than once, the row is the listing whose id is closest to the canonical model id.
ProviderInputOutputCache readCache writeStatus
Kilo Gateway$5.00$30.00$0.50——
LLM Gateway$5.00$30.00$0.50——
LLMTR$5.00$30.00———
Merge Gateway$5.00$30.00$0.50——
NanoGPT$5.00$30.00$0.50——
OpenRouter$5.00$30.00$0.50——
Pioneer$5.00$30.00$0.50$5.00—
Requesty$5.00$30.00$0.50——
Sakana AIreference$5.00$30.00$0.50——
Vercel AI Gateway$5.00$30.00$0.50——
AI/ML API$6.88$41.26$0.69——
EmpirioLabs AI$7.50$45.00$1.50——