Sakana AI发布Fugu Max与Ultra v2多智能体编排模型
Sakana AI Launches Fugu Max and Fugu Ultra v2 for Cheaper, Stronger Multi-Agent Orchestration
多智能体编排是Agent落地关键路径，Sakana用Pareto前沿切分成本与能力场景，Fugu Max的极致低价策略值得开发者关注以优化链路成本。
Sakana AI has released Fugu Max and Fugu Ultra v2, 2 new models in its Sakana Fugu family. Fugu is not a single foundation model. It is a learned orchestrator that routes work across a pool of other models behind 1 API. The new release tunes that architecture for 2 missions. Fugu Max targets the best output per dollar. Fugu Ultra v2 targets the highest capability on hard, multi-step tasks.
Sakana AI 发布了 Fugu Max 和 Fugu Ultra v2，这是其 Sakana Fugu 家族中的两款新模型。Fugu 并非单一的基础模型，而是一个学习型的编排器，它通过一个 API 将工作路由到一组其他模型上。此次新版本针对两项任务对该架构进行了调优。Fugu Max 旨在实现每美元的最佳输出效果。Fugu Ultra v2 则致力于在复杂的多步任务中提供最高的能力。
Is it deployable? Yes, as a hosted API. Both models are live today through Sakana’s OpenAI-compatible API. There are no open weights to self-host, and Sakana does not offer the service in the EU/EEA.
是否可部署？是的，作为托管 API 使用。这两款模型今天已通过 Sakana 的 OpenAI 兼容 API 上线。没有可供自托管的开源权重，且 Sakana 不在欧盟/欧洲经济区（EU/EEA）提供服务。
Why Sakana Frames This as a 2-Axis Problem
为何 Sakana 将其视为双轴问题
Sakana’s argument is direct. Real workloads are judged on capability and cost together. Sending a simple data lookup to a multi-trillion-parameter model wastes money. A better system picks the cheapest machinery that can still solve the task.
Sakana 的观点很直接：实际负载需同时考量能力和成本。将简单的数据查询发送给拥有万亿参数的模型会浪费资金。更好的系统会选择仍能完成任务的最便宜的计算资源。
Sakana team describes this with the Pareto frontier. On that frontier, gaining quality costs more, and cutting cost loses quality. Fugu Max and Fugu Ultra v2 share 1 core orchestration architecture. Only the optimization target differs.
Sakana 团队用帕累托前沿（Pareto frontier）来描述这一点。在该前沿上，提升质量意味着更高的成本，而降低成本则会牺牲质量。Fugu Max 和 Fugu Ultra v2 共享同一个核心编排架构，唯一的区别在于优化目标不同。
The release follows a fast cadence. Fugu entered beta in April, reached general availability in June, and added Fugu-Cyber and a Claude Code interface in July.
此次发布遵循快速的节奏。Fugu 于 4 月进入测试版，6 月实现通用可用性，7 月增加了 Fugu-Cyber 和 Claude Code 界面。
How Fugu Orchestration Works
Fugu 编排的工作原理
The Sakana Fugu’s Technical Report describes Fugu models as language models in their own right. They read a query and build an agentic scaffold for it on the fly. Training combines large-scale fine-tuning, evolutionary algorithms, and reinforcement learning.
Sakana Fugu 的技术报告将 Fugu 模型描述为本身即具备语言模型能力的实体。它们读取查询并即时构建代理式脚手架（agentic scaffold）。训练过程结合了大规模微调、进化算法和强化学习。
The system builds on 2 ICLR 2026 papers. TRINITY uses a lightweight evolved coordinator that assigns Thinker, Worker, or Verifier roles across turns. The Conductor is trained with reinforcement learning to discover natural-language coordination strategies and focused prompts.
该系统建立在两篇 ICLR 2026 论文之上。TRINITY 使用一个轻量级的进化协调器，在多个回合中分配思考者（Thinker）、工作者（Worker）或验证者（Verifier）角色。Conductor 则通过强化学习进行训练，以发现自然语言协调策略和聚焦提示词。
Fugu Max: More Models, Less Cost
Fugu Max：更多模型，更低成本
Fugu Max widens the pool of models Fugu can orchestrate. It adds a large set of open-weights and specialized models. That includes the NVIDIA Nemotron family, through Sakana’s collaboration with NVIDIA. Fugu Max routes each task to the leanest model capable of solving it.
Fugu Max 扩大了 Fugu 可编排的模型池。它添加了大量开源权重和专用模型，其中包括通过与 NVIDIA 的合作引入的 NVIDIA Nemotron 系列。Fugu Max 将每个任务路由给最精简且能解决该任务的模型。
Sakana team reports the following:
Sakana 团队报告如下：
- Pricing: $2 per 1M input tokens and $6 per 1M output tokens.
- Output price: 40% to 60% lower than Sonnet 5, GPT 5.6 Terra, and Kimi K3.
- Performance: Best overall score on 6 benchmarks: Terminal Bench 2.1, GPQA Diamond, AA-LCR, GDP.pdf, AutomationBench, and SWEFish.
- Efficiency: Expands the cost-performance Pareto frontier on 7 of 10 benchmarks.
- 定价：输入令牌每百万 2 美元，输出令牌每百万 6 美元。
- 输出价格比 Sonnet 5、GPT 5.6 Terra 和 Kimi K3 低 40% 至 60%。
- 性能：在 6 项基准测试中总分最高：Terminal Bench 2.1、GPQA Diamond、AA-LCR、GDP.pdf、AutomationBench 和 SWEFish。
- 效率：在10项基准测试中的7项上扩展了性价比帕累托前沿。
Sakana places Fugu Max within striking distance of elite models at 2x to 6x lower cost. SWEFish is an internal Sakana benchmark built from its own coding challenges. Treat that result as a vendor signal.
Sakana将Fugu Max置于以2倍至6倍更低成本接近精英模型的位置。SWEFish是Sakana基于其自身编程挑战构建的内部基准。请将此结果视为供应商信号。
Fugu Ultra v2: Raising the Ceiling
Fugu Ultra v2：提升上限
Fugu Ultra v2 targets complex reasoning, autonomous research, and full-stack software development. Its largest gains appear on sustained reasoning over visual and structured data.
Fugu Ultra v2针对复杂推理、自主研究和全栈软件开发。其最大提升体现在对视觉和结构化数据的持续推理上。
- Chartography (visual reasoning and data interpretation): 48.3, versus 27.3 for Opus 5 and 29.5 for Fable 5.
- DeepSWE (real-world software engineering): 74.3, ahead of models priced 3x to 5x higher per token.
- Breadth: Best or joint-best on 5 of 8 benchmarks: GDP.pdf, Chartography, SWEFish, DeepSWE, and Toolathon.
- Consistency: Top 2 on 7 of 8 benchmarks.
- Chartography（视觉推理与数据解读）：48.3，而Opus 5为27.3，Fable 5为29.5。
- DeepSWE（真实世界软件工程）：74.3，领先于每token价格高3倍至5倍的模型。
- 广度：在8项基准测试中的5项上取得最佳或并列最佳成绩：GDP.pdf、Chartography、SWEFish、DeepSWE和Toolathon。
- 一致性：在8项基准测试中的7项上排名前二。
Fable 5, Fable 5.1, and GPT-6-Astra are not in Fugu Ultra v2’s agent pool. The model’s training cutoff is August 28, 2026. Sakana’s main message is frontier output without dependence on any 1 proprietary model. The research team states that this reduces exposure to vendor lock-in, API revocations, and sudden service cutoffs.
Fable 5、Fable 5.1和GPT-6-Astra不在Fugu Ultra v2的代理池中。该模型的训练截止日期为2026年8月28日。Sakana的主要信息是前沿输出不依赖任何单一专有模型。研究团队指出，这降低了对供应商锁定、API撤销和突然服务中断的暴露风险。
Interactive Explainer
交互式解释器
Key Takeaways
关键要点
- Fugu Max costs $2/$6 per 1M input/output tokens and targets output per dollar.
- Fugu Max posts the best overall score on 6 benchmarks and expands the frontier on 7 of 10.
- Fugu Ultra v2 scores 48.3 on Chartography and 74.3 on DeepSWE.
- Ultra v2 reaches these scores without Fable 5, Fable 5.1, or GPT-6-Astra in its pool.
- Both ship today via an OpenAI-compatible API, with a 1-line switch for existing users.
- Fugu Max每百万输入/输出token成本为2美元/6美元，旨在优化每美元的产出。
- Fugu Max在6项基准测试中取得最佳总分，并在10项中的7项上扩展了前沿。
- Ultra v2在Chartography上得分为48.3，在DeepSWE上得分为74.3。
- Ultra v2在未使用Fable 5、Fable 5.1或GPT-6-Astra的情况下达到这些分数。
- 两者今日均通过兼容OpenAI的API发布，现有用户只需一行代码即可切换。
Check out the Technical details and Project page. Also, feel free to follow us on Twitter and don’t forget to join our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.
查看技术细节和项目页面。此外，欢迎在Twitter上关注我们，别忘了加入我们拥有15万+成员的ML SubReddit并订阅我们的新闻通讯。等等！你在Telegram上吗？现在你也可以加入我们的Telegram群组。
Need to partner with us for promoting your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar etc.? Connect with us
需要与我们合作推广您的GitHub仓库、Hugging Face页面、产品发布或网络研讨会等？请联系我们
The post Sakana AI Launches Fugu Max and Fugu Ultra v2 for Cheaper, Stronger Multi-Agent Orchestration appeared first on MarkTechPost.
本文最初发布于MarkTechPost。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力