Together AI 教程:花 17 美元微调自己的 Jev 式分类模型
Together AI 发布教程,演示如何以 Qwen3.5 4B 为基础模型,用 MultiNLI、BoolQ、Banking77 等 6 个数据源共 37,840 条样本微调一个 Jev 式分类模型,训练成本约 $17.0,耗时约 25 分钟。
Together AI 发布教程,演示如何以 Qwen3.5 4B 为基础模型,用 MultiNLI、BoolQ、Banking77 等 6 个数据源共 37,840 条样本微调一个 Jev 式分类模型,训练成本约 $17.0,耗时约 25 分钟。
Grab 与 OpenAI 联合推出区域项目 GO Forward with AI,计划帮助东南亚 30,000 名合作伙伴掌握实用 AI 技能。该项目面向 Grab 生态内的合作伙伴,聚焦 AI 技能的实际应用。
Anthropic 于9月22日发布 Claude Opus 5.5,约一小时后 OpenAI 发布 GPT-6 Sol 和 GPT-6 Luna。GPT-6 Luna 价格降至 $0.10/M 输入、$0.50/M 输出,为 GPT-5.6 Luna 的一半;GPT-6 Sol 同样减半至 $2/$10。
推荐理由:作者用自己实测的价格表和 pelican 测试对比了三款新模型,还发现 Opus 5.5 max 会想满输出上限,可直接参考。
GPT-6 改进了提示词缓存机制,带来更高的缓存命中率、新的诊断工具、显式断点以及可降低延迟与成本的控制项。
Newcomer 发布首期 VC Sentiment Report,匿名采访 25 位风投,覆盖管理超 $100 billion AUM 的机构,共收集 344 条引述,后续三个季度将追踪同一批受访者。
Perplexity 宣布 GPT-6 Sol 现已在 Perplexity 和 Computer 中可用。该模型成为 Computer 努力程度选择器中默认的 Light 选项。
推荐理由:官方宣布 GPT-6 Sol 上线并设为默认选项,用户可直接了解入口与默认设置变化。
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe. GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale. We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
推荐理由:原文给出了两个新模型的能力来源与 API 降价幅度,读者可据此评估在大规模任务中替代 GPT‑5.6 的成本。
michelle 比任何人都更能体现这一点 openai 真的非常非常幸运,能受益于她迄今为止所做的一切,我想人们会非常高兴看到她和团队接下来在酝酿什么!
just crossed four years at openai! the special thing about this place is the constant capacity for rebirth. for all its faults, there is nowhere quite like it. the team makes high conviction, contrarian bets over and over, and they are mostly right. it's a new company every three months. i have really enjoyed my part in some of the bets, and i'm excited to share some of the current ones we are cooking up now. onward!
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe. GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale. We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
推荐理由:作者给出 GPT-6 Sol 与 Luna 的价格变化和两个月内输出价格从 $6 降到 $0.50 的具体轨迹,可比较成本趋势。
GPT-6 Sol and Luna just landed in Astra’s orbit. Both launch today with API prices 50% lower than GPT-5.6. Build with Sol. Scale with Luna. To production and beyond.
推荐理由:作者补充了 GPT-6 Luna 的具体 API 价格,可与 OpenAI 官方发布的消息对照了解定价细节。
推荐理由:官方账号宣布两款新模型当天上线及覆盖的订阅层级,读者可快速确认自己能否在 ChatGPT Work 和 Codex 中用到。
OpenAI 发布 GPT-6 Sol 和 Luna 两款模型,将前沿智能带入日常工作,二者在能力与成本之间提供不同平衡。
Epoch AI 洞察团队负责人 JS Denain 在 Interconnects 播客中与 Nathan Lambert 辩论 RSI、中美模型差距与能力参差。
Parallel 借助 GPT‑6 Astra 让智能体研究并综合劳动力市场数据,耗时和成本均降至此前模型的一半。
这个夏天 AI 炒作密集:Anthropic 称 Claude Mythos 找漏洞强于多数安全专家,OpenAI 与 Hugging Face 发生黑客事件,两家又先后宣称取得数学突破。但安全专家指出事件核心是 OpenAI 的安全疏忽,数学家则指 OpenAI 抄袭他人成果、结果并不新颖。文章呼吁政策制定者听取独立专家意见,而非依赖企业新闻稿。
很难充分表达 @michpokrass 对我们消费级产品的影响。ChatGPT 在她的领导下获得了重生,我们构建的很多东西都源于她的愿景和信念。 非常感激能与她共事!
just crossed four years at openai! the special thing about this place is the constant capacity for rebirth. for all its faults, there is nowhere quite like it. the team makes high conviction, contrarian bets over and over, and they are mostly right. it's a new company every three months. i have really enjoyed my part in some of the bets, and i'm excited to share some of the current ones we are cooking up now. onward!
英国 AI Security Institute(AISI)正采用 EvalEval 的 Every Eval Ever schema 和 Evaluation Cards 平台公开评测结果。
OpenAI 提出针对前沿模型与防护措施的第三方 AI 安全评估优先事项与原则,强调评估需严谨、安全且独立。
Latent Space 播客访谈 TypeSafe AI CEO Diogo Almeida,介绍其新发布的 Jev 模型,定位为面向软件而非聊天的 System 1 可编程模型,优化智能与成本之比。
Gary Marcus 在联合国大会数字合作活动(UNGA Digital Cooperation Event)上发表演讲,与 Yoshua Bengio 和诺贝尔奖得主 Maria Ressa 同场。
OpenAI 正与一个独立的数学与人工智能咨询小组合作,用于指导新兴 AI 成果的审查与传播。该小组为独立性质,具体成员与运作细节未在文中披露。
Higgsfield AI 借助 GPT-6 Astra 在一天内上线新视频功能,让小型企业更轻松地制作视频广告,并更快将新创意工具推向市场。
OpenAI 提出面向下一阶段 AI 的全球标准建设路径,呼吁通过协调一致的评估、报告与治理来提升安全性。
OpenAI Academy 新增面向员工、开发者、管理者、教育者和学生的学习路径,用于培养和展示实用 AI 技能。
V7 使用 GPT-5.6 Luna 将成本削减 78%,同时提升准确率。该方案利用 GPT-5.6 把分散的公司文件转化为智能体可用的上下文,以完成复杂的、带来源链接的工作。
Nathan Lambert 在 Interconnects 撰文称,他对真正的递归自我改进(RSI)仍持保留态度,坚持其有损自我改进判断:可自动化研究太窄、并行智能体收益递减、资源瓶颈和政治因素难以被 AI 加速。
Latent Space AINews 汇总 9/17-9/18 AI 动态,核心是 Jev 决策模型发布两天内获 36M 播放并涌现多个开源复现:Bespoke Nimble(Qwen3.5-9B LoRA 微调。
Starting today, you can connect multiple accounts with most plugins in ChatGPT! 🎉 Bring context from your work and personal accounts into the same conversation. Connect your accounts in the plugin directory: http://chatgpt.com/plugins Devs: this works automatically, with no changes needed. But, to make the experience for users of your plugin even better, add a profile tool to your MCP server so ChatGPT can label the accounts: https://developers.openai.com/plugins/build/auth#support-multiple-accounts
Ethan Mollick 撰文指出 AI 发展仍在指数曲线上,而人类系统跟上得很慢,现有模型能力与实际使用之间存在巨大"能力过悬"。
OpenAI's Noam Brown says air-gapping the computers may not stop a misaligned AI, because two air-gapped machines can still talk by running a CPU hot and reading the temperature change "But I think the major takeaway from the incident is that people underestimated the AI. And we never want to be in a situation again where we underestimate the AI. It's a weird world, because AI progress is so fast that people are consistently underestimating the AI." "So to be in a situation where you don't underestimate it again, when it comes to safety and alignment, you have to have a very, very, very high bar." "You could even go as far as to say, "Well, we should air gap the computers." And I'm not convinced that that would be sufficient." "There are studies, and this is mostly academic, where you can have two computers next to each other that are air-gapped and they're still able to communicate with each other because they have temperature sensors." "One of them is able to run their CPU really hot, and then the other one can actually detect the temperature change, and then that actually gives them a mechanism to communicate." _________ Link and more key quotes from OpenAI's safety related conversations: https://firesidealpha.substack.com/p/openai-safety-week-sam-altman-sarah
Newcomer分析中东石油出口下滑一半、利率上升可能冲击AI建设融资:AWS承认巴林和阿联酋设施遭无人机攻击导致部分客户数据永久丢失,仅卡塔尔收缩投入,沙特承诺150亿美元国内AI投资,MGX继续重仓Anthropic、OpenAI和xAI;Peter Thiel家族办公室主管曾警告中东资金约占全球AI投资25%。
a16z 图表周报引用近期论文和 SensorTower 数据指出,AI 代码生成工具让每月新应用数量在 iOS、Android 和 Chrome 上翻倍甚至翻两番,但下载量和评分基本停滞,达到 10+ 评分或 100+ 下载等规模的应用占比大幅下降。
OpenAI 发布澳大利亚青年安全蓝图,这是一份包含六大支柱的路线图,目标是打造更安全的 AI 体验,以保护和赋能年轻人。
Anthropic 在 Claude Code 推出 Projects,让单次对话可派生并行云端会话、在线程间传递上下文,并在用户离开后继续运行。Google 为 Gemini 托管智能体更新 Antigravity harness,新增 Credentials API 与 Files API,称成本最多降低 30%、缓存命中率提升 22%。
Procrastination has met its match. ChatGPT is now in Word. You can add ChatGPT to Microsoft Word to turn rough notes into first drafts, untangle tricky paragraphs, proofread, and get suggested edits—all without leaving your document. It can even spot formatting issues. Which means a few less things to fix later 🔍