Fugu v1.0
Public AI IQ model profile with source-backed benchmark results and derived ranking context.
Release and Capabilities
| Public release | 2026-06-22 |
|---|---|
| Modalities | text, image |
Sakana AI hosted API, through subscription or pay-as-you-go access via the fugu endpoint
Publisher-reported capabilities
- Sakana AI develops this multi-model orchestration system in Japan. Results measure the coordinated model pool, not a standalone base model.
- Balances response quality and latency through learned model orchestration.
Limitations and availability caveats
- Relies on external model providers and a changing agent pool.
- Published benchmark evidence is provider-reported and covers the June 2026 system.
- Not currently available in the EU/EEA.
- Token pricing varies with the selected underlying model or highest-tier active agent. A single fixed input/output price is not published.
Official release and model references
Measured Coverage and Estimates
Fugu V1.0 has 4 of 39 tracked IQ benchmark inputs across 2 of 7 dimensions. Raw benchmark tables contain source-backed results only. Derived IQ combines all six dimensions using the shared scoring methodology, including estimates where direct coverage is incomplete.
Among dimensions with direct benchmark coverage, this profile's highest estimated scores are Academic Reasoning (IQ 130) and Programmatic Reasoning (IQ 121). These dimension estimates can include statistical adjustments or predecessor priors; they are not raw benchmark percentages or claims of real-world task success.
Benchmark versions, evaluation harnesses, and coverage affect comparisons; a missing result is not a score of zero. See the scoring methodology for estimation and uncertainty limits.
IQ Dimensions
| Dimension | IQ | Coverage |
|---|---|---|
| Abstract Reasoning | 99 | 0/5 |
| Mathematical Reasoning | 104 | 0/5 |
| Academic Reasoning | 130 | 3/6 |
| Programmatic Reasoning | 121 | 1/6 |
| Computer Use | 109 | 0/5 |
| Reliability | 104 | 0/6 |
| Knowledge Work | 102 | 0/6 |
Benchmark Results
| Benchmark | Score | Dimension |
|---|---|---|
| GPQA Diamond | 95.5 | Academic Reasoning |
| Humanity's Last Exam | 47.2 | Academic Reasoning |
| LiveCodeBench | 92.9 | Programmatic Reasoning |
| SciCode | 60.1 | Academic Reasoning |
Compare
Frequently Asked Questions
- What should I know about Fugu v1.0?
- Sakana AI develops this multi-model orchestration system in Japan. Results measure the coordinated model pool, not a standalone base model. The IQ estimate uses six Sakana-reported June 2026 results across three dimensions, with conservative estimates for missing coverage. These are provider evaluations, not independent Artificial Analysis or Vals measurements. AI IQ labels the original June 2026 Fugu evaluation as Fugu v1.0. Sakana calls the API endpoint fugu; its mutable model pool means later API behavior may differ. Token pricing varies with the selected underlying model or highest-tier active agent. A single fixed input/output price is not published.
- What is the IQ of Fugu v1.0?
- Fugu v1.0 has an estimated IQ of 110, which ranks #73 out of 164 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
- Is Fugu v1.0 open source?
- No — Fugu v1.0 is a proprietary model. Its weights are not publicly available, so it is used through hosted APIs and products.
- When was Fugu v1.0 released?
- Fugu v1.0 was publicly released on June 22, 2026.