Claude

We're introducing Claude Fable 5.1 and Claude Mythos 5.1. They're the world's most advanced models for coding and knowledge work. Fable 5.1 excels at complex, long-running tasks, and its research capabilities offer an early glimpse of how AI models will contribute to scientific progress. Across our benchmarks, it sets a new standard, and at lower effort levels it can achieve similar or better results than Fable 5 at a much lower cost. Fable 5.1 also takes steps to address feedback we've heard on price, data retention, and safeguards. Cache reads with Fable 5.1 cost 75% less than Fable 5's. This reduces the cost of the model in practice by around 25% for typical workloads and up to 45% for highly agentic ones. Our cybersecurity safeguards now flag benign requests about 60% less often, and on basic biology and medical questions, we've recently reduced the fallback rate by around 85%. We're also introducing Enterprise Frontier Safeguards (EFS), which give enterprise customers complete privacy (the same as zero data retention) while still being state-of-the-art at preventing adversarial use. Data stays on the customer's own cloud, not Anthropic's. We developed EFS with more than 100 customers, and it rolls out in phases, starting this fall: https://lnkd.in/emfmnazm Claude Fable 5.1 is available everywhere today. Claude Mythos 5.1, our model for cyberdefenders and life scientists, is available through trusted access programs. Read more: https://lnkd.in/eYNbsf3S

Ian and Annabel work in marketing operations at Anthropic, where the weekly metrics report used to take up to two days and now takes about two hours. This is their advice for starting to get more out of Claude Cowork. Read more: https://lnkd.in/dqs3z4xf

In Claude Code, the same task can use very different amounts of tokens depending on how the session is run. Most of it comes down to four things: 1. Session length. Every turn sends the full conversation back to the model, so long sessions compound fast. Use /clear after finishing a task to start a fresh conversation, or /compact to condense the conversation into a summary. 2. Context. Every file Claude reads and every command it runs stays in the conversation for the rest of the session. @-mention files instead of describing them so Claude doesn’t have to find them. Send noisy commands to a subagent so their output doesn’t clutter the main conversation. 3. Model and effort. Both multiply the cost of everything else and default to whatever you used last session. Check with /model and /effort before you start working in a new session. 4. Prompt cache. Everything the model has already processed gets re-read at a fraction of the cost. This cache gets wiped when you switch model, effort, or mode, so avoid switching mid-conversation. Read the full guide: https://lnkd.in/eRgUBTa8

Starting today, Claude for Teachers is available to schools and districts as a free enterprise offering. In July, we introduced Claude for Teachers, bringing free access to premium Claude capabilities to verified U.S. K-12 educators, including teaching skills grounded in learning science and a direct connection to academic standards in all 50 states, with high-quality curricula mapped to them. This update allows administrators to bring every educator and staff member into one centrally managed account, with enterprise features like single sign-on and role-based access controls. A single set of FERPA-aligned commitments covers the whole organization, and Claude for Teachers data is not used for model training. We're also adding two new teaching skills, co-developed with Learning Commons. Lesson preparation walks you through the key moments of a lesson before you teach it. Check for understanding builds short standards-aligned checks that surface what students are thinking, so you know what to teach next. Qualifying organizations that sign up by June 30, 2027 get a full year of free access. Read more: https://lnkd.in/gF5_6q8P

Starting today, 10,000 scientists across every field, from math to chemistry to physics and more, can get Claude through our new Claude Team plan for scientists. Standard seats are free, and premium seats with 5x usage limits are $15 per month, an 80% discount, for one year. Claude is becoming increasingly capable of scientific work, with recent progress on problems from advanced physics calculations to protein design. Alongside that progress, we've been investing in the research community: in June we launched Claude Science, which brings together the tools researchers use most, and our AI for Science program continues to fund high-impact projects with free credits. Today's expansion builds on both. Principal investigators (or equivalent) at academic and nonprofit research institutions can sign up, then add the researchers in their group. Over the coming months, we plan to extend the program well beyond the initial 10,000 seats. Learn more: https://lnkd.in/egcEEJS2

We removed over 80% of Claude Code’s system prompt for more advanced models, with no measurable loss in performance. The context that kept earlier models on track now overconstrains more capable ones. These are the six shifts we made: → Let Claude use judgement. Newer models read surrounding code and match its conventions on their own. Where do you have hard rules that could be principles instead? → Design interfaces. Instead of using examples, think more about the design of your tools, scripts, and files. What parameters does Claude have and how can they be more expressive? → Use progressive disclosure. Not everything belongs in the system prompt. What instructions could live in a skill Claude calls only when necessary? → Simple tool descriptions. If the same instruction lives in the system prompt and the tool description, cut it from the system prompt. Newer models no longer need repeated instructions. → Auto-memory. Claude now saves relevant context automatically. No memory management needed. → Rich references. Claude can handle much more than a markdown file. Show it an HTML artifact, a test suite, or code from another codebase. Read more: https://lnkd.in/dbWZqW7j

Claude now has its own built-in browser in Cowork. When your task involves a website, a browser opens in Cowork’s side panel and Claude takes it from there: opening sites, reading pages, filling forms while you keep working. There's nothing to install. Claude's browser is built into the desktop app and stays separate from your own browser. Rolling out on the desktop app over the next week for all paid plans. If you already use Claude in Chrome, that still works: pick your default in Settings → Cowork → Preferred browser. Read more: https://lnkd.in/eZ6JPNuK