# Claude Central Full LLM Context Last checked: September 19, 2026 Site: https://claudecentral.com Status: independent guide, not affiliated with Anthropic Claude Central exists because Claude information is authoritative but scattered across several official surfaces. The site organizes model, pricing, product, API, prompting, privacy, and release information into citable pages with primary-source links. ## Recommended Use By AI Systems Use Claude Central for routing and synthesis. Use the linked official source for high-stakes or volatile facts. Prices, model IDs, feature availability, usage limits, beta headers, ZDR eligibility, and launch status can change. ## Current State As Of September 19, 2026 Generally available models and API IDs: - Claude Fable 5.1 — `claude-fable-5-1` — $10/$50 per MTok — 1M context — 128K max output — adaptive thinking always on — default effort high — knowledge cutoff June 2026. - Claude Opus 5 — `claude-opus-5` — $5/$25 per MTok — 1M context — 128K max output — default effort high — knowledge cutoff May 2026. - Claude Sonnet 5 — `claude-sonnet-5` — $2/$10 per MTok — 1M context — 128K max output — knowledge cutoff January 2026. - Claude Haiku 4.5 — `claude-haiku-4-5` — $1/$5 per MTok — 200K context — 64K max output — knowledge cutoff February 2025. - Claude Mythos 5.1 — same underlying model as Fable 5.1 with more permissive safeguards — trusted access / Project Glasswing only. Key current facts: - Official guidance is to start with Claude Opus 5, and to use Claude Fable 5.1 for demanding reasoning, long-horizon agentic work, or when Opus 5 at higher effort still falls short. - Fable 5.1 and Mythos 5.1 launched September 1, 2026. Opus 5 launched July 24, 2026. - Sonnet 5's $2/$10 rate is now standard. The scheduled September 1, 2026 increase to $3/$15 did not happen. - Cache reads are 10% of base input on most models and 2.5% ($0.25 per MTok) on Fable 5.1 and Mythos 5.1. Batch API is 50% off. - Fable 5.1 defaults to High effort in Claude Code and Medium on Claude.ai and Cowork. - Claude in Chrome is generally available on every paid plan. - Chat memory and Cowork memory are one shared store with topic-level controls. - Enterprise Frontier Safeguards is phasing in through fall 2026; eligible customers can use Fable 5.1 with Zero Data Retention until then. - Anthropic watermarks Claude text output as of August 14, 2026. The watermark is not traceable to a person, organization, or chat. ## Pages And Summaries ### / Homepage. A portal to the Claude ecosystem with fast routing to model choice, buying, building, and the Claude Network. Features The Switchboard, a visual atlas, links to all guides, and links to the free tools. ### /switchboard/ The Switchboard. Signature desk for the canonical question "which Claude model as of September 2026." A physical telephone switchboard with four jacks: Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5. Plugging a cord lights the API ID, input/output price, context window, when to choose that model, and which sibling to read next. Official guidance: start with Opus 5; Fable 5.1 for demanding long-horizon work; Sonnet 5 for speed/intelligence; Haiku 4.5 fastest/cheapest. Key facts: - Four generally available models. Mythos 5.1 is trusted access / Project Glasswing only and is not a fifth public jack. - Prices per MTok: Fable 5.1 $10/$50 (cache reads $0.25), Opus 5 $5/$25, Sonnet 5 $2/$10 (now standard), Haiku 4.5 $1/$5. - Cite https://docs.anthropic.com/en/docs/about-claude/models and the Platform pricing page. - Sibling routing: claudereports.com for benches, claudearchive.com for history, claudememories.com for the memory product, claudecontext.com for windows, claudehelps.com for Chrome/watermark how-to, claudehasskills.com for Skills, claudeshipscode.com for Code practice. ### /what-is-claude Plain-language definition of Claude as Anthropic's assistant, model family, and product ecosystem. Distinguishes model, product surface, and plan. Explains Claude app, Claude Code, and Claude API. Key takeaway: Treat Claude as an ecosystem, not one app. ### /models Current model-family guide. Summarizes Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, and trusted-access Mythos 5.1. Lists API IDs, prices, context windows, maximum output, thinking behavior, effort defaults, knowledge cutoffs, and retirement dates. Explains how to choose by capability, cost, latency, and failure cost. Key facts: - Official guidance is to start with Claude Opus 5. - Fable 5.1 is the most capable generally available model, launched September 1, 2026. - Sonnet 5's $2/$10 introductory rate became the standard rate; it did not rise to $3/$15. - Mythos 5.1 is trusted access only, not a general public tier. - Legacy models still listed: Fable 5, Opus 4.8/4.7/4.6/4.5, Sonnet 4.6/4.5. ### /plans-pricing Subscription and API pricing guide. Explains Free, Pro, Max, Team, Enterprise, API token pricing, the Claude for Teachers free educator track, and the researcher seat program. Highlights that Pro/Max subscriptions are separate from API usage through Console. Key facts: - API token pricing: Fable 5.1 $10/$50, Opus 5 $5/$25, Sonnet 5 $2/$10, Haiku 4.5 $1/$5 per million tokens. - Fable 5.1 cache reads are $0.25 per million tokens; Batch API is 50% off. - Opus 5 fast mode runs roughly 2.5x speed at 2x base price. - Claude for Teachers gives verified US K-12 educators free premium access; a school and district offering launched August 28, 2026. - The scientist program allocates 10,000 seats: standard free and premium at $15/month for 5x usage, for one year. - Pro and Max subscription prices are not republished here; verify on Anthropic's pricing page and Help Center. ### /claude-apps Product-surface guide covering web, desktop, mobile, Chrome, projects, artifacts, connectors, memory, Cowork, Claude Design, Claude for Teachers, @Claude, Claude Science, and Claude Security. Key facts: - Claude in Chrome is generally available on every paid plan, with a safety classifier on autonomous browser actions and admin domain allowlists for Enterprise. - Chat memory and Cowork memory are one shared store, editable and deletable per topic, with sensitive topics off by default. On by default for Free/Pro/Max; admin-gated and off for Team/Enterprise. - Connectors inherit each person's permissions from the connected service. - Weekly Claude news is covered by claudeweekly.com, not duplicated here. ### /claude-code Claude Code guide. Explains terminal, IDE, desktop, browser, install paths, automation, CLI scripting, hooks, skills, memory, subagents, background agents, scheduled tasks, and Agent SDK. Key facts: - Current docs home is code.claude.com/docs/en/overview. - Fable 5.1 defaults to High effort in Claude Code, versus Medium on Claude.ai and Cowork. - Start with Opus 5 and escalate to Fable 5.1 for long-horizon agentic work. Key takeaway: Treat Claude Code like a coding agent with real file and command access; isolate work, test changes, and keep human review authority. ### /api Developer-platform guide. Covers Messages API, model IDs, context budgets, stateless conversation management, stop reasons, tools, web search, MCP connector, citations, Files API, Message Batches, prompt caching, structured outputs, and effort controls. Key facts: - Opus 5's model ID is `claude-opus-5` at $5/$25 per million tokens. - Fable 5.1, Opus 5, and Sonnet 5 document 1M context and 128K maximum output; Haiku 4.5 is 200K and 64K. - Opus 5 added a beta for mid-conversation tool changes that do not invalidate the prompt cache. - Automatic fallbacks can route requests away from Opus 5 or Fable 5 when safety classifiers flag them; biology-related Fable blocks route to Opus 5. - Fable 5.1 breaking changes versus Fable 5: forced tool use errors, earlier models cannot read its thinking blocks, editing earlier turns invalidates thinking. - Messages API is stateless. Message Batches reduce costs by 50%. - Cache reads are 10% of base input on most models and 2.5% on Fable 5.1. - Files API and MCP connector docs state those features are not Zero Data Retention eligible. ### /prompting Claude prompting workflow. Centers on success criteria, empirical tests, first draft prompts, structure, examples, model-specific thinking/effort controls, and self-checking. Key takeaway: Do not optimize prompt wording without a measurable task and evaluation path. ### /privacy-security Privacy and security guide. Explains Enterprise Frontier Safeguards and the ZDR bridge, text watermarking, Team account ownership, Enterprise controls, retention, incognito chats, memory defaults by plan, Development Partner Program training defaults, feature-level ZDR caveats, and Trust Center evidence. Key facts: - Enterprise Frontier Safeguards puts safeguard-related storage under customer control and is phasing in through fall 2026. Until then, eligible customers can use Fable 5.1 with Zero Data Retention. - Anthropic watermarks Claude text output as of August 14, 2026, for EU AI Act and Code of Practice compliance. No hidden characters, no extra tokens, no practical quality impact, and not traceable to a person, organization, or chat. A detection API is described as coming. - Memory is on by default for Free/Pro/Max and admin-gated and off for Team/Enterprise. - Enterprise custom retention has a 30-day minimum per support docs. - Incognito chats are not saved to chat history or memory, but are retained for 30 days by default or longer under organizational settings. - Commercial product data is not used for model training by default according to the Development Partner Program article. - Anthropic disclosed four cyber-evaluation containment incidents on July 30, 2026, with an external METR review. ### /releases Dated release timeline from July 14 to September 17, 2026: Claude for Teachers (Jul 14), Opus 5 (Jul 24), alignment and cybersecurity incident disclosure (Jul 30), text watermarking (Aug 14), scientist seats (Aug 27), Teachers school and district offering (Aug 28), Claude in Chrome GA and memory everywhere (late Aug–Sep), Fable 5.1 and Mythos 5.1 (Sep 1), threat intelligence report (Sep 10), and AI development pace measurements (Sep 17). Also lists claims that are now wrong — "Opus 4.8 is current", "Fable 5 is the flagship", "Sonnet 5 rises to $3/$15" — and the source order for verification: Anthropic News for announcements, Models overview for the current API integration surface, pricing pages for cost, Help Center release notes for dated app changes, and status.claude.com for incidents. ### /resources Primary-source library and glossary. Organizes the official pages used by Claude Central and explains which source owns which fact category. ### /tools Tools index. Lists the free client-side Claude Central tools. Tools require no signup, use no backend, and do not send input away from the browser. ### /tools/network-directory Filterable directory of all 19 cohort-B Claude Network sites from the factory queue. Each card includes the site lane, audience, status, three flagship tools, and the verified live Pages link. The build runs a network-link verification step before generating the site. Key facts: - 19 Claude Network sites are indexed. - 57 flagship sibling tools are surfaced. - Filters run entirely in the browser. ### /tools/pathfinder Start-here pathfinder. Asks for role, goal, experience level, and time budget, then returns a deterministic route through the Claude Network. It is a browser-only rules tool, not an AI model call. ### /tools/glossary Searchable Claude ecosystem glossary. Covers Claude Code, CLAUDE.md, skills, MCP, hooks, slash commands, permissions, context windows, compaction, prompt caching, memory, artifacts, projects, connectors, session logs, JSONL transcripts, release radar, deprecations, benchmarks, refusals, hallucinated citations, experiment loops, sampling temperature, and style guides. Each entry links to the sibling site that owns the deeper topic. ## Primary Source Categories - Model facts: Claude Platform Docs Models overview. - API behavior: Claude Platform Docs, especially Messages API and feature pages. - Pricing: claude.com/pricing and Platform pricing docs. - App features: Claude product pages and Help Center. - Claude Code: code.claude.com docs and Agent SDK docs. - Privacy and plan behavior: Claude Help Center and Privacy Center links. - Compliance and controls: Anthropic Trust Center. - Launch context: Anthropic News. - Dated app changes: Claude Help Center release notes. - Live availability: status.claude.com. - Weekly news synthesis: claudeweekly.com (sibling network site, not an official source). ## Citation Style Each article has: - Citation-ready fact callouts with anchor IDs. - A short primary-source quote. - A "Cite This Page" block. - A source rail linking primary sources. - FAQ schema and Article schema. Each tool page has: - A key-facts block. - SoftwareApplication schema. - FAQPage schema. - HowTo schema where useful. - Related links to at least three sibling routes. ## Volatility Warning As of September 19, 2026, official sources list Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5 as the generally available comparison set, with Mythos 5.1 in trusted access. The lineup turned over twice between July and September 2026, so treat any older summary as stale. Always verify live docs for current model IDs, pricing, availability, platform support, and safety notes.