Sakana's Fugu Ultra v2 tops 5 of 8 benchmarks by orchestrating open models, no frontier models
SakanaAILabs · x · 2026-09-12
Alongside the Fugu Ultra v2 launch, Sakana AI published a blog laying out the cost-performance two-axis strategy behind its orchestration systems.
- Benchmarks: best or joint-best on 5/8 benchmarks (DeepSWE 74.3, Chartography 48.3 beating Opus 5 and Fable 5, Toolathon, GDP.pdf, SWEFish) — achieved without Fable 5, Fable 5.1, or GPT-6-Astra in the agent pool
- One architecture, two missions: Fugu Max maximizes output quality at minimum cost, expanding the Pareto frontier; Fugu Ultra v2 pushes peak capability on complex multi-step tasks
- Core thesis: the future belongs to systems that know which machinery to deploy per task, not those that indiscriminately call the largest model; an orchestration layer can rival single frontier models without depending on proprietary ones
- Trajectory: April beta proved multi-agent orchestration works as a unified foundation model; June GA and Ultra v1 showed the layer could match closed frontier models
Note: orchestration tokens are billed as standard input/output, so real cost structure requires careful accounting.
More from coding & agent
- Magic Patterns' first engineer: Cursor in Slack, a Patterson review agent, repo clones over worktrees — Y Combinator Blog · 2026-10-28
- Dev builds AI job-search app for his wife, she lands a job, and he open-sources it — CurieuxExplorer · 2026-10-28
- Supercharging LLM classification with logprobs: soft classifiers you can threshold-tune — JnBrymn · 2026-10-28
- Codex daily build series day 23: skipping iCloud plugins for a reset — dotey · 2026-10-28
- Cloudflare Access adds fully customizable pages with HTML + Liquid templates — dinasaur_404 · 2026-10-28
- Opus retried every session 20.6 million times during a memory leak hunt — johnlindquist · 2026-10-28