Claude (Anthropic)
Anthropic's assistant: the current models (Fable 5.1, Opus 5.5, Sonnet 5.5, Haiku 5.5), what each plan costs, and what shluchim use it for as of October 2026.
On this page
Claude is the assistant made by Anthropic. You use it at claude.ai, in the desktop app for Mac and Windows, or in the phone app. In the group it is the usual pick for long documents, Word and Excel files, shiur handouts, grant writing, and anything that involves building a website or a small app. Its two standing complaints are usage limits and the lack of a picture generator.
Anthropic's agent features (Cowork, Dispatch, Claude Design, Claude in Chrome, Claude Tag, Claude Code) have their own page: Claude Cowork and Anthropic's agent tools.
The models, October 2026
Anthropic ships four sizes. A higher number is a newer model. All four below were released in September and October 2026.
| Model | Released | Use it for |
|---|---|---|
| Haiku 5.5 | Oct 7, 2026 | Quick, simple jobs: sorting a list, short replies. Members found older Haiku versions sloppy on anything with judgment. |
| Sonnet 5.5 | Sep 28, 2026 | Everyday writing: emails, posts, letters, summaries. |
| Opus 5.5 | Sep 22, 2026 | Long documents, careful reasoning, building sites and apps, anything that has to hold together across many pages. Anthropic says it performs at the level of Fable 5.1 on most work and costs about 40% less to run than Opus 5. |
| Fable 5.1 | Sep 15, 2026 | The strongest model for the hardest, longest jobs. It burns through usage fast, and on the $20 Pro plan it runs on paid usage credits. |
A sister model, Mythos 5.1, is the same model as Fable with fewer safety limits. It is only open to approved cybersecurity and life-sciences organizations, so it is not something a Chabad House will use.
Fable has strict safety filters. In July 2026 members saw Claude flag ordinary coding work and drop to an older Opus model mid-chat. Anthropic's own announcements confirm that flagged cybersecurity requests fall back to Opus 4.8.
The group's rule of thumb since June 2026: Sonnet for plain writing, Opus for complex work, Fable only when the job is long and has to stay consistent. If you are unsure, ask Claude which model and effort level fit the task before you start.
For how the lineup got here, see The Model Leapfrog.
What it costs
Prices checked on claude.com/pricing on 2026-10-21.
| Plan | Price | What you get |
|---|---|---|
| Free | $0 | Chat, file creation, Claude Design, slides and docs. No Claude Code, no Claude in Chrome, no Opus. |
| Pro | $20/month, or $17/month billed yearly ($200) | Opus, Claude Code, Claude in Chrome, Projects, more usage. Fable only through usage credits. |
| Max 5x | $100/month | Five times Pro usage, Fable up to 50% of weekly limits, $100/month in API credits. |
| Max 20x | $200/month | Twenty times Pro usage, Fable up to 50% of weekly limits, $200/month in API credits. |
| Team Standard | $25/seat monthly, $20/seat yearly, 2 to 150 seats | Claude Code, Cowork, shared admin and billing. |
| Team Premium | $125/seat monthly, $100/seat yearly | Five times a Standard seat's usage, Fable up to 50% of weekly limits. |
Every plan has limits: a five-hour window and, on paid plans, a weekly cap. When you hit one, Claude pauses until it resets unless you buy extra usage. See Paid vs. Free for how the plans compare with ChatGPT and Gemini.
Nonprofit pricing. A registered 501(c)(3) can buy Team seats at $8 per user per month (Standard) or $40 (Premium), with a two-seat minimum, after a few-minute check through Goodstack. A Premium nonprofit seat is the group's favorite way to get Max-level usage at a lower price. Details and gotchas are on Nonprofit Discounts.
Teachers. Anthropic gives verified US K-12 educators a free year of Claude if they sign up by June 30, 2027 (announcement).
What shluchim use it for
These come from the group's chats from August 2025 to October 2026. They are members' experiences, not guarantees.
- Shiur handouts and source sheets. Paste five or six Hebrew quotes and ask for an English class handout; members said even free Claude lays this out well (October 2026). Ask for LaTeX or Typst when the layout must be exact.
- Word and Excel files. Upload a Word template with your styles and tell Claude to follow them exactly; without a template its formatting is plain. Members merged four years of accounting exports into one comparison spreadsheet (February 2026).
- Grant writing and letters. A grant writer said Claude frames a concept paper and grant request better than ChatGPT (July 2026). Others prefer it for pastoral and thank-you letters. See Letter Writing.
- A study partner. One shliach uses Opus as a chavrusa that tracks the shitos and open questions in a sugya across sessions and pushes back (June 2026). Check every source it gives; see AI and Torah Accuracy.
- A secretary. A shliach keeps one running Claude thread of recurring programs, holiday plans, vendors and volunteers, tied to Google Calendar (October 2026).
- Websites and small apps. Members built event pages, seat-reservation pages, a donor plaque page and shul magazines. "Opus 5.5 plus reference images" was the October 2026 advice for an elegant site from scratch. See Vibe Coding.
- CRM work. Members connect Claude to Salesforce (and so to ChabadOne CRM) through a connector and say it works well once you explain the data model first (October 2026). See CRM Automation.
Where it falls short
- No picture generator. As of October 2026 members still report that Claude does not make photos or illustrations. The usual workaround is to ask Claude to write the image prompt, then run it in ChatGPT or Gemini. See Image Generation.
- Usage limits. This was the top complaint through spring 2026. On May 6, 2026 Anthropic doubled Claude Code's five-hour limits on paid plans and dropped peak-hour cuts for Pro and Max, after a compute deal with SpaceX (announcement). Complaints came back in late August and September 2026 as members moved to Fable 5 and then Fable 5.1, which use far more of the allowance.
- Copyright refusals. Members reported Claude refusing to transcribe a PDF until they confirmed they owned it, and a scheduled task that would not pull a parsha summary from Chabad.org (August and October 2026). See AI Refusals and Workarounds.
- Long chats drift. In a long project Claude can forget a page size or a format it followed earlier. Save each finished piece, and start a fresh chat when it goes off course.
- Same-looking designs. Members noticed Claude's sites all use similar fonts and colors. Give it reference images or a style guide, or use Claude Design.
- Published artifacts are public. If you publish an artifact to get an embed link for ChabadOne, anyone with the link can see it (members, September 2025).
Habits that help
Turn a task you repeat into a skill: in a new chat type /skill-creator, describe the task, answer its questions, and save the skill under Settings. A Project is a workspace with standing instructions and files; a skill is a set of instructions Claude can pull into any chat.
The group's long-running "one step at a time" prompt still works for anything technical:
When I ask how to do something, work with me step by step, ONE step at a time.
Don't give me all the steps at once. Start with a short explanation of what
we're going to do. Then give me only the first step. Wait for me to confirm it
worked. Then move to step 2.
To connect GitHub: Settings, then Connectors, then GitHub. Pick the repository and start prompting.
Running an agent on a Claude subscription
Members who run a personal agent (see GabAI and Hermes Agent) power it with a Claude subscription instead of paying per token. In May 2026 Anthropic announced that this kind of use (the Agent SDK, claude -p, and third-party apps built on it) would move to a separate monthly credit on June 15. On June 15 it paused the change. Anthropic's help page still says you can use the Agent SDK and third-party apps with your subscription limits (support article). Since October 7, 2026, Max and Team plans also include monthly API credits. The full story, and the AWS billing trap that caught one shliach, is on Paid vs. Free.
What changed in 2026
- February: Opus 4.6 (Feb 5) and Sonnet 4.6 (Feb 17).
- March: Members noticed that Claude for Nonprofits now took teams of two instead of five (two is the published minimum today). Dispatch (Mar 23) let you send Cowork tasks from your phone.
- April: Opus 4.7 (Apr 16) and Claude Design (Apr 17).
- May: Higher limits after the SpaceX deal (May 6). Opus 4.8 (May 28).
- June to July: Fable 5 launched June 9, was suspended for all users on June 12 when US export controls were applied, and returned July 1. Sonnet 5 (June 30) became the Free and Pro default. Claude Tag brought Claude into Slack (June 23). Opus 5 followed on July 24.
- September to October: Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. The Opus 5.5 launch came with higher five-hour limits and a one-time limit reset subscribers could save for later.