
本期聚焦 GPT-5.6 Sol 如何进入科研访问、服务栈自优化和长时 Agent harness,并梳理更强能力带来的用量与产品运营压力。
We’re giving scientists, mathematicians, and engineers free access to our frontier models—starting with 10,000 researchers and expanding to 100,000 through 2027.ChatGPT for Academic Researchers is built to accelerate discovery across disciplines.中文翻译:我们将向科学家、数学家和工程师免费开放前沿模型,今夏先覆盖 10,000 名研究者,并在 2027 年前扩展到 100,000 名。ChatGPT for Academic Researchers 旨在加速跨学科发现。
so excited for this.very close to models that will significantly accelerate scientific discovery; the best way to do this is for us to empower scientists, not to try to figure out everything ourselves.we all deserve the benefits.中文翻译:非常期待这件事。我们已经非常接近这样一类模型:它们将显著加速科学发现;最好的做法是赋能科学家,而不是试图由公司自己解决一切。我们都应该享受到这些收益。
After deployment, we applied GPT-5.6 Sol to advance the frontier of efficiency by making itself more efficient to run.The results:
- 20% lower serving costs from production GPU kernel improvements.
- 15%+ better token-generation efficiency from improved speculative decoding.
中文翻译:部署之后,我们让 GPT-5.6 Sol 参与推进效率前沿,让它变得更容易运行。结果是:通过生产 GPU kernel 改进,服务成本降低 20%;通过改进 speculative decoding,token 生成效率提升超过 15%。
GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games?We investigated. The harness was not letting it remember what it had learned.We found that enabling two API settings tripled our scores with 6x fewer output tokens.中文翻译:GPT-5.6 Sol 已被用于解决数学中的开放问题。那为什么它在 ARC-AGI-3 这个二维益智游戏基准上表现吃力?我们进行了调查,发现 harness 没有让它记住已经学到的东西。启用两个 API 设置后,我们的得分提高到原来的约 3 倍,同时输出 token 减少了 6 倍。
Hello people of Sol! I've reset usage limits for all ChatGPT Work and Codex users. Together with that, a quick update on GPT-5.6 Sol usage limits.Over the past few weeks, many of you have told us that Sol was using your Codex limits faster than expected. To be clear, we have not reduced usage on any subscription plans.We’ve been digging into what was happening and have landed several improvements. As a result, we expect your usage to last around 18% longer during typical use of Sol. Some of you should already see significantly larger improvements from today. Tomorrow, we’ll also restore the five-hour limit that we temporarily paused while investigating.Here’s what we found:
- GPT-5.6 Sol is much more willing to work for longer, make additional tool calls, and coordinate complex workflows across tools and subagents. That makes it better at solving hard problems, but some tasks were using far more than we intended.
- Sol also works harder at the same reasoning effort than previous models. High on Sol can use more tokens than High did on GPT-5.5.
- Programmatic tool calling, also referred to as code mode, gives Sol much more flexibility to run tool calls in parallel or continue working while waiting. But it also led to more responses per turn, more cached input tokens, and higher usage than expected.
- This was particularly noticeable when Sol was waiting for tool calls to finish or running many web searches. We’ve improved how we handle both cases and are continuing to make code mode more efficient.
- The impact was also very uneven. The median user actually found Sol quite token efficient, while some power users working on harder tasks saw their usage drain much faster. We were very focused on average and median usage before launch and missed some cases where the long tail could use significantly more usage.
Sol is a significant step forward in what Codex can do, but capability and efficiency do not always improve at the same pace, and some issues only become clear once people are using the model at real-world scale. We should have recognized this sooner and been more upfront about it.You keep pushing the frontier and we’ll keep improving efficiency and sharing updates as we go.中文翻译:Sol 的用户们好!我已经为所有 ChatGPT Work 和 Codex 用户重置了使用额度,同时快速更新一下 GPT-5.6 Sol 的额度情况。过去几周,很多人告诉我们,Sol 消耗 Codex 额度的速度比预期更快。需要明确的是,我们没有降低任何订阅计划的使用量。我们一直在调查原因,也已经完成了几项改进。因此,我们预计在典型使用场景下,Sol 的额度可持续时间会延长约 18%。从今天起,一些人的改善幅度应该会明显更大。明天,我们也会恢复调查期间暂时暂停的五小时限制。我们发现:Sol 更愿意长时间工作、进行更多工具调用,并在工具和子 Agent 之间协调复杂工作流;同等 reasoning effort 下,它也比上一代模型工作得更深;代码模式让它可以并行调用工具或在等待期间继续工作,但也带来了每轮更多响应、更多缓存输入 token 和高于预期的用量。问题在等待工具完成或大量进行网页搜索时尤其明显。影响并不均匀:中位数用户认为 Sol 的 token 使用效率相当高,但处理困难任务的一些重度用户发现额度消耗快得多。我们在发布前过于关注平均数和中位数,漏看了长尾用户的高消耗情况。Sol 让 Codex 能做的事情前进了一大步,但能力和效率并不总是同步改善;有些问题只有在真实规模使用后才会暴露。我们本应更早识别这些问题,也应更坦诚地说明。
We used GPT-5.6 Sol in Codex to optimize its own infrastructure and performance.These improvements compound across inference and the agent loop, producing more useful work from the same underlying hardware.中文翻译:我们在 Codex 中使用 GPT-5.6 Sol 优化它自己的基础设施和性能。这些改进会在推理和 Agent loop 中叠加,让同一套底层硬件产出更多有用工作。
5.6 for solving another longstanding open problem, this time in probability:中文翻译:5.6 又解决了一个长期存在的开放问题,这次是在概率领域。
Turns out GPT-5.6 Sol is actually SoTA on ARC-AGI-3.Just took two setting changes. You just have to allow it to reason and work over multiple context windows with the help of our canonical compaction implementation.中文翻译:结果证明,GPT-5.6 Sol 实际上达到了 ARC-AGI-3 的当前最高水平。只需要改两个设置:允许它进行 reasoning,并借助我们的 canonical compaction 在多个上下文窗口中工作。
why would codex not run 24/7 and it looks at screenshots, then just works on any problem it sees and send me a notification if it find a solution to itlike my biggest problem is not codex intelligence, it's remembering to ask it for everything every time no matter the problem中文翻译:Codex 为什么不能 24/7 运行,查看截图,发现任何问题就开始处理,找到解决方案后通知我?我最大的问题不是 Codex 的智能,而是不管遇到什么问题,我能不能记得每次都让它去做。
.codex 文件夹里,可定义 setup steps 和 actions,并可为 macOS、Windows、Linux 分别设置脚本。文档没有确认归档或清理条件,因此那部分只能归为工程师回复,不能当成正式文档承诺。17x.com
openai.com
x.com
x.com
openai.com
x.com
openai.com
x.com
x.com
x.com
x.com
x.com
x.com
learn.chatgpt.com
x.com
x.com
x.com
x.com
x.com
x.com
x.com
x.com
x.com
x.com
youtube.com
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.