Signature experience · September 19, 2026
The Switchboard
As of September 19, 2026, four Claude models are generally available: Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5. Official guidance: start with Opus 5; use Fable 5.1 for demanding long-horizon work; Sonnet 5 for speed and intelligence; Haiku 4.5 for the fastest, cheapest tier. Plug a jack for API ID, price, and context. Claude Central is independent and not affiliated with Anthropic.
Line of September 2026
Claude Central operator desk
Lift a cord. Plug a jack.
Four generally available models. The lamp lights the API ID, the token price, the window, and the sibling to read next. There is no fifth public jack.
Keys 1–4 plug a jack when the cabinet has focus. Arrow keys walk the row. The board ships with Opus 5 already patched, because that is the official start.
Patch to the rest of the network
This desk answers which model. It does not steal sibling pages.
Independent analysis of vendor numbers. This desk does not republish benchmark tables as measured fact.
Historyclaudearchive.comThe dated timeline: what shipped, what retired, what a current page should not rewrite.
Memory productclaudememories.comChat and Cowork memory, topic controls, and who can turn it on. Not the context window.
Windowsclaudecontext.com1M and 200K windows, compaction, and prompt-cache economics, including Fable 5.1 cache reads at $0.25.
Chrome / watermarkclaudehelps.comEveryday how-to: Claude in Chrome GA, and what text watermarking is and is not.
Skillsclaudehasskills.comWhat a Skill is, where the directory lives, and how to author one. Code install belongs on Ships Code.
Code practiceclaudeshipscode.comClaude Code as practiced: defaults, effort, hooks, and the start-with-Opus-5 path in a repo.
Four Jacks, Not Five
The public comparison set on September 19, 2026 is Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5. Anthropic’s models overview is the source for names, API IDs, and windows. The Platform pricing page is the source for token prices. The choosing-a-model page is the source for the order: start with Opus 5.
Mythos 5.1 exists. It is the same underlying model as Fable 5.1 with more permissive safeguards, the same $10/$50 token price, and a trusted-access / Project Glasswing gate. It is not a generally available fifth jack, and this board will not pretend it is. If a guide still leads with Opus 4.8, Fable 5, or a Sonnet 5 rise to $3/$15, it is describing July, not September.
How To Read A Lit Jack
API ID first. The string you send is claude-opus-5, claude-fable-5-1, claude-sonnet-5, or claude-haiku-4-5. Haiku also publishes a dated alias, claude-haiku-4-5-20251001. Do not invent a slug.
Price is input then output, per million tokens. Fable 5.1 is $10/$50. Opus 5 is $5/$25. Sonnet 5 is $2/$10, now the standard rate. Haiku 4.5 is $1/$5. Cache reads are 10% of base input on most models and $0.25 on Fable 5.1 (and on Mythos 5.1, which is not on this board). The Batch API is 50% off. Subscription seat prices are not restated here; verify those on Anthropic.
Context is not the memory product. Fable, Opus, and Sonnet document a 1M window and 128K max output. Haiku documents 200K and 64K. Windows, compaction, and cache-read math belong on Claude Context. The product that remembers you across chat and Cowork belongs on Claude Memories.
Benches are not this page. Anthropic published Fable 5.1 figures in the September 1 announcement. This desk will not launder vendor numbers into independent measurement. Claude Reports owns that analysis. Dated history that should not be rewritten lives on Claude Archive.
The rest of the cord rack is similarly fenced. Chrome and watermarking how-to: Claude Helps. What a Skill is: Claude Has Skills. Claude Code as practiced, including the High-effort Fable default inside Code: Claude Ships Code. The written model guide on this site remains /models.
FAQ
Which Claude model should I use in September 2026?
Start with Claude Opus 5 (`claude-opus-5`, $5/$25 per million tokens, 1M context). Use Claude Fable 5.1 for demanding reasoning and long-horizon agentic work, Claude Sonnet 5 when you want the speed/intelligence balance at $2/$10, and Claude Haiku 4.5 when you want the fastest, cheapest tier. Anthropic documents that lineup on the models overview. Mythos 5.1 is trusted access / Project Glasswing only, not a fifth generally available model.
What is Claude Fable 5.1 vs Opus 5 vs Sonnet 5 vs Haiku 4.5?
Fable 5.1 is the most capable generally available model ($10/$50, 1M context, cache reads $0.25). Opus 5 is the official starting point ($5/$25, 1M). Sonnet 5 is the balance point, and its $2/$10 rate is now standard. Haiku 4.5 is fastest and cheapest ($1/$5, 200K context). Those four are the public comparison set as of September 19, 2026.
Did Sonnet 5 pricing increase on September 1, 2026?
No. The $2 input / $10 output per million tokens rate announced as introductory through August 31, 2026 became the standard rate. It did not rise to $3/$15. Verify on the Platform pricing page.
Is Claude Mythos 5.1 a generally available model?
No. Mythos 5.1 is the same underlying model as Fable 5.1 with more permissive safeguards, priced $10/$50, and limited to trusted access / Project Glasswing. The Switchboard has four jacks because there are four generally available models. It does not invent a fifth.
Is Claude Central affiliated with Anthropic?
No. Claude Central is an independent editorial publication. It is not affiliated with Anthropic. Plug the official models overview and pricing pages before you ship, buy, or bill.
Cite this page
Claude Central. “The Switchboard: which Claude model as of September 2026.” Last checked September 19, 2026. https://claudecentral.com/switchboard/
The Switchboard is original work by Claude Central. It is an editorial instrument, not an Anthropic product. Claude Central is an independent publication and is not affiliated with Anthropic.