OpenAI 连发 GPT-5.6 Sol 与语音模型
OpenAI 宣布 GPT-5.6 Sol 及 Terra、Luna 三模型本周四公测，同时上线 GPT-Live 新一代语音交互模型。对工程师而言，多模型分级策略可能改变 API 选型逻辑，语音实时交互则打开了新的产品形态。
Sam Altman 只发了一个"GPT"就获得 13 万赞，这种极简发布策略本身就是在制造 FOMO，与 GPT-4 发布前的预热套路一致。三模型命名（Sol/Terra/Luna）暗示了明确的能力分级，可能是 OpenAI 首次公开采用 MoE 架构的不同规模变体，而非简单的 speed/tier 区分。
LeCun 的转发内容与此完全无关，说明他选择在这个时间点刻意淡化 OpenAI 发布，这种"不回应"本身就是回应——Meta 的 Llama 4 语音能力可能也在同一窗口期。做语音 AI 应用的团队需要同时盯紧两边的 API 定价，OpenAI 的语音模型如果按分钟计费，成本结构会和 ElevenLabs 直接竞争。
GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday. We’re expanding preview access globally now. https://t.co/Uk5HcfSc2e
查看原文 →RT @JaredRyanSears: For anyone who is worried about the communist threat facing America, here is some important information: No one was ea…
查看原文 →RT @ddale8: Greenland isn’t surrounded by Chinese and Russian ships. There hasn’t been $19.2 trillion invested in the US under Trump. Trump…
查看原文 →Introducing GPT-Live, a new generation of voice models for natural human-AI interaction. Rolling out in ChatGPT starting today. You’ll want to turn the sound on for this one. https://t.co/WzoQFvA5ir
查看原文 →GPT-live (next-generation voice) launches today in ChatGPT. it feels magical and 'real'. i have always preferred typing to talking to an AI, now i think that's going to shift.
查看原文 →RT @HQNewsNow: 1,200 former DOJ employees have just sent a letter urging Senators not to confirm Todd Blanche as AG: "In the coming weeks,…
查看原文 →RT @Microinteracti1: For the past year, MAGA has been screaming “Europoor” like a cheap car alarm going off at three in the morning. Nobody…
查看原文 →RT @ylecun: @ZhugeEX The *median* wealth of a French is almost *twice* the median wealth of an American. So much for the "Europoor" meme.…
查看原文 →RT @SteveRattner: The U.S. trade deficit rose by over 40% last month, reaching $77.6 billion. That's the largest since March 2025 and high…
查看原文 →The next generation of ChatGPT Voice is here. Livestream starts at 10am PT. https://t.co/giS4K1yNh9
查看原文 →RT @SamanthaJPower: One year since USAID was dismantled, and the bill is coming due. On @TheLastWord last night, Lawrence O'Donnell and I t…
查看原文 →RT @sapinker: The data are in: Populism makes countries worse off. "... a country being ruled by a populist is associated with a 10% lower…
查看原文 →RT @joshwoodward: What's something that you're surprised @GeminiApp can't do well, and we should have fixed a long time ago?
查看原文 →RT @RemiCadene: Starting with the fundamentals Prototype Version 0 AI, Software, Hardware A small team, 9 months Designed and assembled i…
查看原文 →RT @RemiCadene: Unveiling our Northstar design @UMA_Robots A body that can navigate our space Hands that can use our tools Approachable,…
查看原文 →RT @GoogleDeepMind: As @Apptronik expands their Robot Park facility, our research partnership means real-world data collected by the latest…
查看原文 →RT @DarrigoMelanie: More than half of Americans can’t afford food or gas. Meanwhile, the top .1% increased their wealth by $2.59T during T…
查看原文 →I've already talked to two companies that limit engineers to $100 in weekly tokens. That's around $20,000 in extra costs per person per year. After you run out, you gotta write the code yourself, like a caveman. I suspect most companies will end up here. There's no way they can offer an open buffet of tokens without going bankrupt.
查看原文 →How you can build a moat with self-learning agents: If you can build an agent that gets better every time people use it, you will be unstoppable. Here is what you can do: 1. Learn from two sources, not one. Agent traces show what the agent did and where it broke. In-browser activity shows how users steered and fixed the results. Most products capture the first source and forget the second. 2. You have three options to apply new learnings: fine-tune your model, update the harness, or provide in-context information to the agent. Make sure you take advantage of all three. 3. When deciding what your agent should remember, lean on procedural memory (where you store workflows and rules) and episodic memory (where you store specific things that happened). Don't over-index on semantic memory, as it can go stale and become a liability. 4. Set boundaries around how far learning spreads (per-user, per-team, per-app). Be careful not to leak facts between different users' contexts. 5. You want to own the data your agent learns from. This data is your most valuable asset. Try to keep it in your infrastructure so it doesn't get stolen by Big Cloud. 6. Use open, framework-agnostic standards to capture and apply learnings. Here is an article with more information:
查看原文 →RT @rohanpaul_ai: During a Bloomberg interview, Yann LeCun (@ylecun ) explains why LLMs are limited in terms of real-world intelligence dur…
查看原文 →RT @mcgovernmit: McGovern neuroscientists have discovered that you don't need language to think logically. People with severe language impa…
查看原文 →RT @ylecun: @steeve @medjawii @videolan @FFmpeg Super-fun in-depth discussion 🙏🙏🙏 World models, JEPA vs generative architectures, anti-coll…
查看原文 →RT @JeanRemiKing: Excellent investigation of the mechanics of language models. 🫢A notable blindspot, however: several teams have actually…
查看原文 →RT @steeve: We're releasing ZML/LLMD, our homegrown LLM server built on top of our homegrown high performance heterogeneous inference stack…
查看原文 →RT @LarryBoorstein: DOGE shut down USAID, established by JFK in 1961. Before being shut down, the US Agency for International Development…
查看原文 →RT @RemiCadene: AI developed at @UMA_Robots Single neural network, End to End It operates robustly for hours, with precision, adapting to…
查看原文 →RT @steeve: We sat down with @ylecun for a 1h30 discussion on world models @medjawii and JB Kempf of @videolan/@FFmpeg. It's probably the…
查看原文 →RT @SuryaGanguli: Check out all the amazing work from our @SimonsFdn Collaboration on the Physics of Learning and Neural Computation (https…
查看原文 →RT @psermanet: At UMA, we build and own the full stack, from hardware to software. This allows us to bake safety in at all levels, rather t…
查看原文 →An email for AI agents is actually an interesting idea. I guess it’s like having an employee with its own email address and you cc’d them whenever you need their help. This will keep separation between your email and the agent’s. It will also allow the agent to act freely on whatever emails they get without any concerns.
查看原文 →