#13Sakana· released 2026-06-24

Sakana: Fugu Ultra.

Fugu Ultra is the higher-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to route...

View on OpenRouter ↗
Context
1.0M
Input $/1M
$5.00
Output $/1M
$30.00
Modality
text · image
Arena Elo
1553
LMArena — human blind votes
AA Index
64.0
Artificial Analysis composite
LiveBench
82.2
Contamination-resistant
GPQA
85.6%
Grad-level science reasoning
MMLU-Pro
84.7%
Broad knowledge
SWE-bench
71.9%
Real GitHub bugs (Verified)
Terminal-Bench
60.4%
Agentic shell tasks
ARC-AGI
20.1%
Abstract reasoning
Aider Polyglot
77.9%
Multi-language code editing
Speed
39 tok/s
Median output speed

Profile

Sakana: Fugu Ultra is Sakana's entry at position #13 on the current aggregated leaderboard. Its strongest signals are graduate-level science reasoning, real-world software engineering, unattended agentic work, and very long context. Every figure below is aggregated from the public sources listed on the homepage and refreshed on each pipeline run.

The score spread is comparatively complete for this model, but the standing advice still applies: treat gaps of a few points as ties and validate on your own workload before committing.

How this model's position is computed is documented in the methodology; what each benchmark actually measures is in the benchmark guide. See a figure that disagrees with its source? Report it — accepted corrections apply on the next refresh.

Nearby on the leaderboard