Disclosure: This article was written with AI assistance and reviewed by EmpirioLabs AI.
Fugu Ultra v1.1 is available on EmpirioLabs. It is Sakana AI's updated multi-agent conductor for difficult reasoning, coding, and research tasks. One request coordinates a deeper pool of expert agents and returns their combined work as one answer.
What changed in v1.1
Fugu Ultra v1.1 adds a distinct max reasoning level for the hardest problems. The high level balances speed and performance, xhigh provides deep reasoning for complex problems, and max is the distinct maximum level. Reasoning remains always on.
Version ids are explicit on EmpirioLabs. Use fugu-ultra-v1-1 for the new model and fugu-ultra-v1-0 for the original release. Existing integrations that send fugu-ultra remain pinned to v1.0, so adopting v1.1 is an intentional model change rather than a silent upgrade.
How to call Fugu Ultra v1.1
Use the OpenAI-compatible chat completions endpoint and set model to fugu-ultra-v1-1:
curl https://api.empiriolabs.ai/v1/chat/completions \
-H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
-H 'Content-Type: application/json' \
-d '{
"model": "fugu-ultra-v1-1",
"reasoning_effort": "max",
"max_tokens": 4096,
"messages": [
{
"role": "user",
"content": "Design a migration plan for a multi-region API with zero downtime."
}
]
}'The same model also works through the Responses, Messages, and Google-compatible content-generation endpoints. Output limits must be at least 16 tokens, and complex prompts benefit from generous client timeouts.
Capabilities to build with
Fugu Ultra v1.1 accepts text and image input within a 1M token context window. It supports function calling, strict JSON Schema structured output, and built-in web search. Web search can cite sources when the response provides them.
The model performs orchestration before it answers, so requests can take from a few seconds to a few minutes. Streaming requests stay open, but the answer is delivered when orchestration finishes rather than token by token.
Pricing and billing
Fugu Ultra v1.1 uses pay-as-you-go, context-tiered token pricing with discounted cached input. Billing includes the prompt, final answer, and the orchestration tokens used by the expert-agent pool. Larger contexts use the higher context tier. See the live Fugu Ultra v1.1 model page and pricing page for current rates.
Start building
Try Fugu Ultra v1.1 in the Playground, read the API documentation, or compare both versions in the model catalog.
