以下翻译只覆盖原文的前一部分,其余内容请看下方原文。
← AI资讯AI资讯
2026年11月17日 —— GPT-6.2 Astra 和 Sol 降低了基准测试成本。早期测试者对使用 Astra 的体验意见不一。
OpenAI 今天发布了 GPT-6.2 Astra 和 Sol,这比 Anthropic 预计的首个交易日早两天。在 Artificial Analysis 的测试中,Sol 与九月份的 Astra 持平,每项任务的成本约为 15%。新版 Astra 在智能指数(Intelligence Index)上得分为 62,略低于 Fable 5.2 的 65 分。总体而言,Anthropic 仍然拥有全球最佳模型,但 OpenAI 拥有全球最具性价比(每美元智力产出最高)的模型。

这次发布更不寻常之处在于影响者(KOL)给出了强烈的负面反馈……

我们认为其中一些反应是过度反应,这是影响者的正常倾向,要么觉得一切令人惊叹 🤯🤯🤯,要么就是简直是史上最烂.MODEL.EVER,但也有一些批评是合理的,强化学习(RL)产生的伪影明显且简直 bizarre(甚至别提那个糟糕的发布视频了)。不过,我们仍然用它来自动化我们的工作,这才是最终重要的事。
发布得很好,但 Anthropic 的 IPO 不受影响,而这可能将是本周/今年/本十年(直到 OpenAI IPO……)的头版新闻。
今天的 AI 资讯由 SerpApi 赞助!
通过 URL 检索公共网页,格式为 LLM 可用的 Markdown 或结构化 JSON。使用一个 API 映射网站并运行受限爬取。配合 100 多个 SerpApi 搜索 API,实现从搜索到内容的流程。
AI Twitter 回顾
GPT-6.2 Astra 和 Sol
-
可用性。 今日发布的模型正在向 ChatGPT Work 以及 Plus、Pro、Enterprise 和 Edu 用户的 Codex 推出,并通过 API 向开发者提供 gpt-6.2-astra 和 gpt-6.2-sol。OpenAI
-
API 每百万 token 价格。 Astra 输入价格为 $8,缓存输入为 $0.40,输出为 $40。Sol 保持其前身定价不变,分别为 $2、$0.10 和 $10。两者均具有 1M 上下文窗口和 128K 输出限制。Vals · Ultrafast 定价和发布
-
Astra 用于法律领域。 Enterprise 配置增加了法律搜索索引。Artificial Analysis 在 Harvey LAB-AA 上的测评为 20.8%,而基础版 Astra 为 15.0%。Artificial Analysis
-
准备情况。 OpenAI 将 Astra 在 AI 自我改进方面的评级定为 High,这是其外部发布模型中的首次。在另一项指标上,Vals 的 RSI-Index 从 27.07% 上升至 29.01%,其中 50% 代表人类参考水平。系统卡片 · Vals
-
FrontierMath Erdős。 Epoch 验证了 Lean 中 68 个问题中的七个解决方案。之前的最佳成绩是三个。Epoch
基准测试衡量什么
-
Agents’ Last Exam 1.1。 OpenAI 报告称,Astra 在 xhigh 设置下比 GPT-6 Astra 在 max 设置下高出 5.2 分,且每项任务的成本不到三分之一。在该测试中,它比 Fable 5.2 领先 5.5 分,成本仅为后者的约 5%。在 GDPval v2 上,它表示 Astra 比其前身少使用了 22% 的输出 token。OpenAI 的结果
-
长工程运行。 Mercor 的 OpenWorldEngineer 显示 Astra 为 67.2%,Fable 5.2 为 54.6%,GPT-6 Astra 为 44.2%。在 Vals 的多轮 Vibe Code Bench 上,Astra 排名第一,Sol 排名第三。Mercor · Vals
-
3D 和机器人技术。 Astra 以 1,318 分领先 AA-3D,高于 Fable 的 1,271 分和 Sol 的 1,259 分。在 Artificial Analysis 的机器人预览中,Astra 得分为 79,Sol 为 78,高于 Fable 的 75 分。Artificial Analysis
-
研究代理。 在 Perplexity 的 WANDR 上,Astra 得分为 0.783,每项任务成本约为 $12.54。Sol 达到 0.746,每项任务成本约为 $2.65。Perplexity
早期测试者意见不一
-
Matt Shumer 在经过数周测试后称这些模型“基本上不可用”,并表示他的 Opus 和 Fable 代理拒绝了 Astra 子代理的大部分工作。评论 · 工作流详情
-
Pietro Schirano 将他早期的许多问题归因于测试框架(harness)。Arena 的 Peter Gostev 将 Astra 与 Fable 在视觉编码方面进行了对比,并单独指出了 Astra 生成的 SVG 孔雀。Schirano · Gostev
-
演示文稿。 Gabriel Chua 使用超过 100 个 Codex 线程构建了一个可交互的新加坡。Simon Willison 将相同的 three.js 机甲提示词输入到多个模型中,包括 Astra、Sol、Carbon 和 Fable。新加坡构建 · 同提示词对比
Microsoft Ignite 和 Gemini
-
微软发布自己的编码模型。 MAI Code 2 用于多步编码,Code 2 Flash 用于更快的编辑。两者均可在 GitHub Copilot 的模型选择器、Copilot Fusion 和 Foundry 中使用。微软还宣布对其 Decision 模型进行多模态升级。Microsoft
-
Copilot Fusion。 微软报告称,在其多模型框架下,Terminal-Bench 4.1 得分达 71.2%,成本比该比较中的 Claude Fable 5.2 低 88%。Jay Parikh
-
图像编辑。 Artificial Analysis 将其编辑基准测试的第一名授予 MAI Image 2.7,每 1,000 张图像成本为 $34.82。Artificial Analysis
-
Gemini Tasks。 Hydrogen 现在在六个国家提供免费 Tasks 服务,包括需要电话或短信的任务。Google 还展示了一个带有 MuJoCo 模拟的新 Antigravity 机器人插件。Sundar Pichai · Google DeepMind
代理框架和推理
Hugging Face 的 Pi 适配器。 在 pi 内部训练的小型适配器将 Qwen3.8-27B 在 SWE-bench Pro 最难的 51 个任务中的解题数从 15 提升到 32。在该子集上,每项任务的工具调用减少了 55%。模型卡片和配对结果
-
框架成本差异巨大。 Baseten 在其包含 22 个框架和五个开源模型的基准测试中,发现通过任务的单位成本差异高达 4.4 倍。Baseten
-
框架可能会针对其模型专门化。 Google 的 SIM 保留先前框架编辑的理由和结果。其在 Opus 上进化的框架在 Gemini 上得分为 68.41,而初始框架得分为 73.89。研究摘要
-
更多多代理产品。 Cognition 推出了持久化的 Devin Bots。Perplexity 报告称,其混合模型团队在 Terminal-Bench 和 WANDR 上击败了主模型的四个实例。Replit 报告称,在 NumPy 任务中,代理数量从 16 增加到 32 时出现了收益递减现象。Cognition · Perplexity · Replit
-
协调主导了 1,000 个代理的运行。 Hugging Face 表示,在任何时刻只有 40–60 个代理在调用模型。它们每写入一个 token 大约读取 130 个 token,合并和测试导致了瓶颈。Thomas Wolf
-
一小部分缓存未命中可能主导账单。 SemiAnalysis 报告称,在其样本中,1.8% 的未缓存请求约占总支出的 15%。SemiAnalysis
-
Marin 和 K2 Horizon 公开了他们的训练过程。 Marin 发布了其 535B-A23B 运行中的 57 个中间 Mochi 检查点。K2 报告比较了在相同 22T token 上训练的四种模型大小,并描述了一种解码适配器,据称可实现 2.2 倍的加速。Marin · K2 Horizon
-
Zyphra 的长上下文注意力。 该团队报告称,在将几个模型转换为其二次以下复杂度注意力机制后,在 10M token 处实现了 9–60 倍的加速,Qwen 在 1M 处的 RULER 测试中损失不到两个百分点。Zyphra
-
推测性解码。 vLLM 报告称,对于 Kimi K3.1 Pro,滑动窗口草稿使其接受长度提高了多达 30%。NVIDIA 也为 Qwen3.8-27B 提供了一个 4B 的草稿模型。vLLM · 草稿模型研究
行业和政策
-
CEO 交接后的 NVIDIA 首份财报。 在临时 CEO Colette Kress 的带领下,NVIDIA 季度营收达 1,158 亿美元,同比增长 103%,并表示 Vera Rubin NVL72 已发货。董事会表示无意恢复 Jensen Huang 的职位。NVIDIA · AP
-
Isomorphic Labs 融资 50 亿美元。 Greenoaks 和 Sequoia 联合领投了 C 轮融资。该公司表示将扩大其药物设计平台规模并开发治疗管线。Isomorphic Labs
-
Genesis AI 融资 5 亿美元。 A 轮融资使这家机器人公司的投后估值达到 35 亿美元。它计划利用这笔资金部署其 Eno 机器人并扩展培训和数据收集。Zhou Xian
AI Reddit 回顾
本地模型和硬件
-
一个具有双层运行的 9B 模型。 SenseNova-Loop-9B 将其 32 层中的 24 层运行两次。模型作者报告称,GAIA 分数从预览版的 37.86 上升到发布版的 69.90。本地用户询问在长上下文情况下 KV 缓存还剩多少内存。权重 · 讨论
-
Maxwell V1 Flash。 这个 35B/3B-active MoE 声称在 256K 上下文下,其稀疏注意力比密集 GQA 快 6.7 倍。它仍然保留完整历史的 KV 缓存,这是本地用户在承诺购买硬件前希望衡量的特性。模型卡片 · 讨论
-
Olala mini 7A1B。 OVHcloud
查看原文
← AI NewsAI News
November 17, 2026 — GPT-6.2 Astra and Sol lower benchmark costs. Early testers disagree on what Astra is like to work with.
OpenAI shipped GPT-6.2 Astra and Sol today, two days before Anthropic’s expected first trading day. In Artificial Analysis’ tests, Sol ties September’s Astra at about 15% of the cost per task. New Astra scores 62 on the Intelligence Index, just behind Fable 5.2 at 65. Net net, Anthropic still owns the world’s best model, but OpenAI owns the world’s most efficient models in terms of intelligence per $.

What’s more unusual about this launch is the STRONGLY negative influencer feedback…

We think that some of this is overreacting and normal influencer tendency to either have everything be 🤯🤯🤯 or the literal WORST.MODEL.EVER, BUT also that some of the criticisms are valid, and the RL artefacts are noticeable and downright bizarre (don’t even get us started about the terrible launch video). However, we still use it to automate our work and that’s all that matters in the end.
Great ship, but Anthropic’s IPO is unaffected and that will be probably the headline of the week/year/decade (until OpenAI IPO’s….)
Today’s AI News is brought to you by SerpApi!
Retrieve public web pages by URL as LLM-ready Markdown or structured JSON. Map sites and run bounded crawls with one API. Pair with 100+ SerpApi search APIs for search-to-content workflows.
AI Twitter Recap
GPT-6.2 Astra and Sol
-
Availability. Today’s models are rolling out to ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, and to developers through the API as gpt-6.2-astra and gpt-6.2-sol. OpenAI
-
API prices per million tokens. Astra is $8 input, $0.40 cached input and $40 output. Sol keeps its predecessor’s pricing at $2, $0.10 and $10. Both have 1M context and a 128K output limit. Vals · Ultrafast pricing and rollout
-
Astra for Law. The Enterprise configuration adds a legal search index. Artificial Analysis measured 20.8% on Harvey LAB-AA, versus 15.0% for base Astra. Artificial Analysis
-
Preparedness. OpenAI rated Astra High for AI Self-Improvement, a first for one of its externally released models. On a separate measure, Vals’ RSI-Index rose from 27.07% to 29.01%, with 50% denoting its human reference. System card · Vals
-
FrontierMath Erdős. Epoch verified seven solutions in Lean out of 68 problems. The previous best was three. Epoch
What the benchmarks measure
-
Agents’ Last Exam 1.1. OpenAI reports Astra at xhigh gained 5.2 points over GPT-6 Astra at max, using less than a third of the cost per task. It reports a 5.5-point lead over Fable 5.2 at about 5% of the cost on that test. On GDPval v2, it says Astra used 22% fewer output tokens than its predecessor. OpenAI’s results
-
Long engineering runs. Mercor’s OpenWorldEngineer puts Astra at 67.2%, Fable 5.2 at 54.6%, and GPT-6 Astra at 44.2%. On Vals’ multi-turn Vibe Code Bench, Astra was first and Sol was third. Mercor · Vals
-
3D and robotics. Astra led AA-3D with 1,318, ahead of Fable at 1,271 and Sol at 1,259. On Artificial Analysis’s robotics preview, Astra scored 79 and Sol 78, ahead of Fable at 75. Artificial Analysis
-
Research agents. On Perplexity’s WANDR, Astra scored 0.783 at about $12.54 per task. Sol reached 0.746 at about $2.65. Perplexity
Early testers disagreed
-
Matt Shumer called the models “essentially unusable” after weeks of testing and said his Opus and Fable agents rejected much of Astra’s subagent work. Review · Workflow details
-
Pietro Schirano attributed many of his early issues to the harness. Arena’s Peter Gostev tested Astra against Fable on visual coding and singled out Astra’s SVG peacock. Schirano · Gostev
-
The demos. Gabriel Chua built a walkable Singapore using more than 100 Codex threads. Simon Willison ran the same three.js mech prompt through several models, including Astra, Sol, Carbon and Fable. Singapore build · Same-prompt comparison
Microsoft Ignite and Gemini
-
Microsoft ships its own coding models. MAI Code 2 is for multi-step coding, and Code 2 Flash is for quicker edits. Both are available in GitHub Copilot’s model picker, Copilot Fusion and Foundry. Microsoft also announced a multimodal upgrade to its Decision model. Microsoft
-
Copilot Fusion. Microsoft reports 71.2% on Terminal-Bench 4.1 with its multi-model harness, at 88% lower cost than Claude Fable 5.2 in that comparison. Jay Parikh
-
Image editing. Artificial Analysis ranks MAI Image 2.7 first on its editing benchmark, at $34.82 per 1,000 images. Artificial Analysis
-
Gemini Tasks. Hydrogen now powers free Tasks in six countries, including tasks that need a phone call or a text. Google also showed a new Antigravity robotics plugin with MuJoCo simulation. Sundar Pichai · Google DeepMind
Agent harnesses and inference
Hugging Face’s Pi adapter. A small adapter trained inside pi lifted Qwen3.8-27B from 15 to 32 solves on the 51 hardest SWE-bench Pro tasks. Across that subset it made 55% fewer tool calls per task. Model card and paired results
-
Harness costs vary sharply. Baseten measured up to a 4.4× spread in cost per passing task across its benchmark of 22 harnesses and five open models. Baseten
-
The harness may specialize to its model. Google’s SIM keeps the rationale and outcome of prior harness edits. A harness it evolved on Opus scored 68.41 on Gemini, versus 73.89 for the starting harness. Research summary
-
More multi-agent products. Cognition launched persistent Devin Bots. Perplexity reported that its mixed-model teams beat four instances of the lead model on Terminal-Bench and WANDR. Replit reported diminishing returns from 16 to 32 agents on NumPy tasks. Cognition · Perplexity · Replit
-
Coordination dominated a 1,000-agent run. Hugging Face says only 40–60 agents were calling the model at any moment. They read roughly 130 tokens for every token they wrote, with merges and tests contributing to the bottleneck. Thomas Wolf
-
A small fraction of cache misses can dominate the bill. SemiAnalysis reports that 1.8% of uncached requests accounted for roughly 15% of spend in its sample. SemiAnalysis
-
Marin and K2 Horizon opened up their training runs. Marin released 57 intermediate Mochi checkpoints from its 535B-A23B run. The K2 report compares four model sizes trained on the same 22T tokens and describes a decoding adapter that the team says reaches 2.2× speedup. Marin · K2 Horizon
-
Zyphra’s long-context attention. The team reports 9–60× speedups at 10M tokens after converting several models to its subquadratic attention, with Qwen losing under two points on RULER at 1M. Zyphra
-
Speculative decoding. vLLM reports up to 30% higher acceptance length from a sliding-window draft for Kimi K3.1 Pro. NVIDIA also has a 4B draft for Qwen3.8-27B. vLLM · Draft-model research
Industry and policy
-
NVIDIA’s first earnings after the CEO transition. Under interim CEO Colette Kress, NVIDIA posted $115.8B in quarterly revenue, up 103% year over year, and said Vera Rubin NVL72 had shipped. The board said it did not intend to reinstate Jensen Huang. NVIDIA · AP
-
Isomorphic Labs raises $5B. Greenoaks and Sequoia co-led the Series C. The company says it will scale its drug-design platform and develop its therapeutic pipeline. Isomorphic Labs
-
Genesis AI raises $500M. The Series A values the robotics company at $3.5B post-money. It plans to use the funding to deploy its Eno robot and expand training and data collection. Zhou Xian
AI Reddit Recap
Local models and hardware
-
A 9B model with layers that run twice. SenseNova-Loop-9B runs 24 of its 32 layers a second time. The model’s authors report GAIA rising from 37.86 in the preview to 69.90 in the release. Local users asked how much memory remains for the KV cache at long context. Weights · Discussion
-
Maxwell V1 Flash. The 35B/3B-active MoE claims a 6.7× attention speedup over dense GQA at 256K. Its sparse attention still retains the full-history KV cache, which is what local users want measured before they commit hardware. Model card · Discussion
-
Olala mini 7A1B. OVHcloud previewed a small MoE with 1B active parameters and a 32K context window. The thread is waiting for weights and reproducible benchmarks. OVHcloud · Discussion
-
Intel’s Crescent Island numbers. The early brief’s eye-catching throughput comparison uses FP64-heavy workloads. Intel’s separate LLM comparison claims 1.12× throughput and 29% lower latency versus RTX PRO 6000 Blackwell Server Edition. Hardware discussion
-
Gemma on a printer. A builder reports running Gemma 3 270M via a PostScript print job. Demo
Agents in use
-
A video-game agent built for real-time input. miHoYo’s Gamer samples the screen at 5 Hz and emits keyboard and mouse activity at 30 Hz. Its RL version solved about half of the held-out Honkai puzzle tasks in the authors’ evaluation. Project · Discussion
-
A fictional league runs a nightly broadcast. The builder says Fock Ball simulates the games in a physics engine and uses Claude for surrounding work such as commentary and the newsroom. Builder’s post
-
ICLR reviews and AI detectors. Reviewers report that the program chairs flagged likely AI-generated reviews. The thread asks how a detector can distinguish editing help from whole-review generation. Peer review discussion
Thanks for reading
Stay a little ahead of the curve.
Fresh ideas, original reporting, and deep conversations in your inbox.
Subscribe to Latent Space
原文版权归 Latent Space 及原作者所有,本站转载仅供阅读。如需删除,请联系 support@aisa.one。