No Fable 5, No Problem: Sakana AI Says Fugu Hits the Same Marks
Sakana AI has launched Fugu, a multi-agent orchestration system, with its flagship Fugu Ultra claiming to match the efficiency of Anthropic’s export-restricted Fable and Mythos fashions.
The Tokyo-based startup positions Fugu as a hedge in opposition to vendor lock-in, arguing that orchestrating a pool of swappable fashions lets customers sidestep the restrictions that just lately pulled Fable and Mythos offline.
How Sakana Fugu Works and Why It Matters
Fugu is a big language mannequin (LLM) that calls different fashions in an agent pool. It may even invoke itself recursively. Users ship one request, and the system handles choice, delegation, and synthesis internally.
The product ships in two tiers through a single OpenAI-compatible API.
- Fugu targets on a regular basis work with decrease latency.
- The flagship mannequin Fugu Ultra coordinates a bigger agent pool for extra complicated duties, comparable to AI analysis and cybersecurity evaluation.
The launch follows Anthropic’s June 12 suspension of access to Fable 5 and Mythos 5. A US export management directive cited nationwide safety and barred entry for overseas nationals.
Sakana’s pitch faucets the sovereign AI debate, the concept that nations want control over the fashions they depend upon.
The crew argued that any authorities or enterprise that depends on a single supplier’s AI for important methods “is a cloth vulnerability.” Sakana mentioned that the menace has shifted from theory to something already enjoying out.
“As now we have seen with latest export controls imposed on fashions like Fable and Mythos, entry can disappear in a single day. Collective intelligence is the sensible hedge in opposition to this focus of energy. Because Fugu orchestrates an underlying pool of swappable brokers, it merely routes round vendor restrictions. By orchestrating the world’s fashions, we’re delivering the resilient blueprint required for true AI sovereignty,” the crew mentioned.
Follow us on X to get the newest information because it occurs
Meanwhile, progress can be being made in different nations. India’s Sarvam AI recently reached a valuation of $1.5 billion.
How Fugu Ultra Stacks Up Against Fable, Mythos, and Frontier Rivals
Sakana benchmarked Fugu in opposition to Fable 5, Mythos Preview, Gemini 3.1 Pro, GPT 5.5, and Opus 4.8. The outcomes are blended slightly than a clear sweep. Fugu Ultra led 4 of eight exams, together with LiveCodeBench at 93.2 and GPQA-D at 95.5.
Fable 5 came out on top in three benchmarks. It topped SWEBench Pro at 80.0, properly above Fugu Ultra’s 73.7. Fable 5 additionally led on SciCode and Humanity’s Last Exam, whereas Opus 4.8 edged forward on CTI-REALM.
Against Mythos Preview, Fugu Ultra gained every shared benchmark, however solely by sub-point margins. Beyond the charts, Sakana additionally ran job demonstrations.
In an autonomous machine studying check, Fugu Ultra ran greater than 100 experiments over 14 hours on a single H100 GPU.
“Fugu Ultra (daring crimson) completed with the greatest imply efficiency (0.9774). Fugu Ultra additionally achieved the greatest single run of the whole experiment (0.9748), main each single baseline,” the crew mentioned.
The different demos lined buying and selling, chess, and design. Overall, Fugu joins a rising effort by nations and companies to construct alternate options as entry to frontier fashions tightens.
Subscribe to our YouTube channel to observe leaders and journalists present skilled insights
The publish No Fable 5, No Problem: Sakana AI Says Fugu Hits the Same Marks appeared first on BeInCrypto.
