[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nvidia-nooa-python-agent-framework":3,"news-related-32b938b6-01a3-43c9-b040-14db6c5f57c6":38},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"32b938b6-01a3-43c9-b040-14db6c5f57c6","NVIDIA 把 Agent 装进一个 Python 类:被忽略的 NOOA,一半 token 跑出 SWE-bench 82.2%","AVO 满分抢走头条的同时,NVIDIA 7 月底开源的 Python Agent 框架 NOOA 被大多数人跳过:整个 Agent 塌缩成一个类,docstring 即提示词、注解即契约,SWE-bench Verified 82.2%,token 消耗约为同类框架一半——harness 正在追上模型本身。","过去两周 NVIDIA 的 AI 新闻几乎全被 AVO 刷屏:8 月 22 日,这套 Agent 系统在 ARC-AGI-3 上拿到满分 100,清掉全部 183 个关卡,不少人把它当成\"模型又变强了\"的证据。但同一家公司 7 月 30 日悄悄开源的另一个项目——Python Agent 框架 NOOA(NVIDIA Object-Oriented Agents),被大多数人直接跳过了。这两件事其实指向同一个判断:模型之上的执行层,正在成为能力提升的主战场。\n\n## NOOA 是什么:把 Agent 折进一个 Python 类\n\nNOOA 的全称是 NVIDIA-labs Object Oriented Agents,Apache 2.0 许可证,`pip install nooa` 即装。它的核心设计可以用一句话讲完:Agent 不是一堆配置文件和回调,Agent 就是一个 Python 类。类上的字段是状态,普通方法是确定性的能力,docstring 就是提示词,类型注解就是输入输出契约;而那些函数体只写 `...` 的方法,会在运行时交给 LLM 驱动的循环去实现。模型是可插拔的——通过 LiteLLM 支持 Anthropic、OpenAI、Ollama、vLLM 等主流后端,换底座不用改 Agent 代码。\n\n这个设计的直接收益是工程性的:Agent 可以像普通 Python 对象一样被测试、追踪、重构和版本管理。对任何一个维护过 prompt 目录加工具 schema 加回调配置三件套的团队来说,这个减法很有吸引力。\n\n## 82.2% 的另一面:token 账单砍半\n\n成绩单上最硬的数字是 SWE-bench Verified 82.2%。第三方报道给出的对照更能说明问题:达到这个分数,NOOA 每个任务平均消耗约 110 万 token、调用约 28 次模型,而同类框架普遍在 220 万 token、66 次调用左右——大约是前者一半的量级。marktechpost 与 aiweekly 等多家媒体的报道相互印证了这一组数字。另外据 aitoolsrecap 的评测记录,NOOA 还报告了 CyberGym L1 86.8% 和 ARC-AGI-3 85.1% 的成绩;不过这些是厂商自报数字,且框架尚处 v0.0.8 alpha 研究预览阶段,读的时候留一分保留。\n\n值得强调的是,82.2% 是 NOOA 框架搭配 GPT-5.5 跑出来的。也就是说,同样的模型,换一个组织方式更干净的 harness,分数和成本可以同时改善——这正是 AVO 满分故事的同款逻辑:模型没换,换的是外面包的那一层。\n\n## 为什么\"抽象\"值得当新闻看\n\nNOOA 值得关注的点不在跑分,而在它把行业里一个越来越响的共识摆上了台面:Agent 的可靠性瓶颈,很大一部分不在模型,在接口组织方式。现有框架把提示词、工具、回调、工作流拆成四个独立抽象,每多一层胶水就多一分出错面;NOOA 把它们压回 Python 原生的类语义里,让\"写 Agent\"回归到\"写软件\"。GitHub 上 1.9k star 的早期热度说明社区对这个方向有真实兴趣。\n\n当然要泼一盆冷水:README 明确警告 NOOA 是研究软件,Agent 会执行 LLM 生成的代码,进程内的 AST 校验和模块黑名单只是防御纵深,真正的隔离边界必须靠容器、VM 这类操作系统级沙箱。想上生产的话,先读安全章节。\n\n## 给读者的\"所以呢\"\n\n接下来选型的时侯,与其问\"哪个模型最强\",不如先问\"我的 harness 配不配得上我的模型\"。NOOA、AVO 这一轮演示的共同结论是:同一个模型,套上不同质量的执行框架,能差出一条及格线。框架层开源、可审计、可替换,它值得拿到和模型选型同等的注意力——这可能是近期最被低估的一条选型线索。","https:\u002F\u002Fgithub.com\u002FNVIDIA-NeMo\u002Flabs-OO-Agents","998df6db-96e6-4b8e-8be1-cfa00a6cd177",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":19,"name":20,"slug":20,"description":14,"color":14},"8dac812d-3839-4abe-a855-5f56ec9515fd","nvidia",{"id":22,"name":23,"slug":23,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"3c23e5e6-7e96-490f-a073-52f3e578d010","en","NVIDIA NOOA: The Open-Source Python Agent Framework That Skipped the Headlines","While NVIDIA's AVO flooded headlines with a perfect ARC-AGI-3 score, the company quietly open-sourced NOOA (NVIDIA Object-Oriented Agents) on July 30: an Apache 2.0 Python framework that collapses an entire agent into a single class, hitting 82.2% on SWE-bench Verified at roughly half the token cost of peer frameworks.","NVIDIA's AI headlines of the past two weeks belonged almost entirely to AVO: on August 22 the agent system posted a perfect 100 on ARC-AGI-3, clearing all 183 levels, and much of the coverage framed it as proof that \"the model got smarter again.\" But another project the same company quietly open-sourced on July 30 — the Python agent framework NOOA (NVIDIA Object-Oriented Agents) — was skipped by most observers. Both events point to the same conclusion: the execution layer above the model is becoming the main battleground for capability gains.\n\n## What NOOA Is: An Agent Folded Into One Python Class\n\nNOOA's full name is NVIDIA-labs Object Oriented Agents. It ships under Apache 2.0 and installs with `pip install nooa`. Its core design fits in one sentence: an agent is not a pile of config files and callbacks — an agent is a Python class. Fields on the class are state, ordinary methods are deterministic capabilities, docstrings are the prompts, and type annotations are the input-output contracts. Methods whose bodies contain only `...` are implemented at runtime by an LLM-driven loop. The model is pluggable: through LiteLLM it supports Anthropic, OpenAI, Ollama, vLLM and other mainstream backends, so swapping the base model requires no changes to agent code.\n\nThe immediate payoff is engineering hygiene: agents can be tested, traced, refactored, and version-controlled like ordinary Python objects. For any team that has maintained the classic three-piece set of a prompt directory plus tool schemas plus callback configs, this subtraction is genuinely attractive.\n\n## The Other Side of 82.2%: Half the Token Bill\n\nThe hardest number on the scoreboard is 82.2% on SWE-bench Verified. The comparison reported by third-party coverage makes the point better: to reach that score, NOOA consumes roughly 1.1 million tokens and about 28 model calls per task, while peer frameworks typically run around 2.2 million tokens and 66 calls — roughly double on both counts. Reports from marktechpost and aiweekly corroborate this set of numbers. According to aitoolsrecap's review notes, NOOA also reports 86.8% on CyberGym L1 and 85.1% on ARC-AGI-3; these are vendor-reported figures, and the framework is still a v0.0.8 alpha research preview, so read them with a grain of salt.\n\nWorth emphasizing: the 82.2% was achieved by NOOA paired with GPT-5.5. In other words, the same model, wrapped in a harness with a cleaner organization, improved on score and cost at the same time — which is exactly the logic of the AVO story: the model did not change; what changed is the layer wrapped around it.\n\n## Why an \"Abstraction\" Deserves News Coverage\n\nThe interesting thing about NOOA is not the score but the industry consensus it puts on the table: a large share of the reliability bottleneck in agents lies not in the model but in how interfaces are organized. Existing frameworks split prompts, tools, callbacks, and workflows into four separate abstractions, and every extra layer of glue adds surface area for errors. NOOA compresses them back into native Python class semantics, turning \"writing an agent\" back into \"writing software.\" An early 1.9k stars on GitHub suggests real community interest in this direction.\n\nA bucket of cold water, though: the README explicitly warns that NOOA is research software, that agents execute LLM-generated code, and that in-process AST checks and module deny-lists are defense-in-depth only — the real containment boundary must come from OS-level sandboxes such as containers or VMs. Before taking it to production, read the safety section first.\n\n## The \"So What\" for Readers\n\nWhen making your next stack decision, instead of asking \"which model is strongest,\" ask first \"whether my harness deserves my model.\" The shared lesson of NOOA and AVO is that the same model, wrapped in execution frameworks of different quality, can differ by a passing grade. The framework layer is open source, auditable, and swappable — it deserves the same attention as model selection, and it may be the most underrated selection signal of the season.","nvidia-nooa-python-agent-framework","2026-08-23T17:20:00Z","2026-08-23T17:15:16.677380Z","2026-08-23T17:15:16.677388Z",true,"agent",35,{"items":39},[40,45,50,55,60,65],{"id":41,"title":42,"news_slug":43,"published_at":44},"0237222a-602b-47ef-9431-468009904428","FACET 先建环境再写任务:1.2K 轨迹把 Qwen3.5-27B 推到 Terminal-Bench 47.57,逼近 397B","facet-terminal-task-synthesis","2026-08-19T06:19:20+00:00",{"id":46,"title":47,"news_slug":48,"published_at":49},"4bb93e6c-a626-4e8e-b5f3-3ea3a15c8604","NVIDIA Molt 把 Agentic RL 训练栈打回 PyTorch 原生:一个异步循环训多模态 MoE,代码小到 LLM 一次读完","nvidia-molt-agentic-rl-pytorch","2026-07-28T03:00:00+00:00",{"id":51,"title":52,"news_slug":53,"published_at":54},"56cb62a1-da4f-4ac6-94ee-e60346f8d075","英伟达 BioNeMo Agent Toolkit：生命科学库塞进 AI Agent","nvidia-bionemo-agent-toolkit-life-science","2026-06-24T00:00:00+00:00",{"id":56,"title":57,"news_slug":58,"published_at":59},"e3d0e837-43fc-49a6-b533-fbb75aff95d3","FastContext 4B-30B 子代理：把 60% 推理 Token 留给\"找代码\"，Coding Agent 提速 5.5%","fastcontext-msr-sub-agent-code-search-5-5pct","2026-06-20T02:01:00+00:00",{"id":61,"title":62,"news_slug":63,"published_at":64},"ad3e5dbd-2c30-43a1-bf67-a6ccd16fa11e","Databricks 开源 Omnigent：Matei Zaharia 想给 Coding Agent 之上再加一层「元 Harness」","databricks-omnigent-meta-harness-coding","2026-06-13T08:00:00+00:00",{"id":66,"title":67,"news_slug":68,"published_at":69},"f26ace13-9c96-47ea-a528-b6682a22aa1e","Apodex 1.1 把推理搬进真实执行:PIVOT-RL 定位关键决策点,35B mini 开源","apodex-1-1-agentic-execution-pivot-rl","2026-08-25T14:30:00+00:00"]