Claude 5: Everything Anthropic Has Shipped in 2026

The wait for “Claude 5” is over. Anthropic never shipped one monolithic model called Claude 5, so the fifth generation arrived tier by tier instead, and on July 24, 2026 the last gap closed when Claude Opus 5 replaced Opus 4.8 as the flagship at $5 per million input tokens and $25 per million output. Anthropic has kept moving since: Claude Fable 5.1 and Claude Mythos 5.1 arrived on September 1, 2026, and on September 22 Claude Opus 5.5 opened a 5.5 family at $4/$20 and pushed Opus 5 onto Anthropic’s legacy list.

That makes the honest answer to “is Claude 5 out” a straight yes across every tier, and the generation has already been added to twice. This guide lays out exactly what shipped and when, what each model is actually for, how the flagship compares to the Mythos class above it, and which Claude to use today. No leaked screenshots, no fake login pages, just the confirmed lineup and where it leaves you.

The Key Takeaways

• There is no single “Claude 5” model. Anthropic versions each tier on its own, so the 5-generation landed in pieces rather than at one launch.
• Claude Opus 5 shipped July 24, 2026 at $5/$25 per million tokens under the API id claude-opus-5, closing the last gap in the generation.
• It went to number one on Artificial Analysis at launch, and it is not there now. On the current Intelligence Index v4.3.2 it reads 50.8, behind Claude Opus 5.5 on 57.6, Claude Fable 5.1 on 53.4 and GPT-6 Astra on 52.7.
• Claude Sonnet 5 shipped June 30, 2026 at $2/$10, an introductory rate made permanent on Aug 10, 2026, with a 1M-token context window.
• The Mythos tier still matters. Fable 5 and its unrestricted sibling Mythos 5 launched June 9, and Mythos 5 remains well ahead of the Opus line on offensive security work.
• The generation has moved twice since. Claude Fable 5.1 and Claude Mythos 5.1 shipped on September 1, 2026, and Claude Opus 5.5 opened a 5.5 family on September 22 at $4/$20, taking the top of the Intelligence Index at 57.6.

Is Claude 5 Out? Yes, and the Last Gap Just Closed

Anthropic does not release a numbered “Claude 5” the way OpenAI ships a “GPT-5.” Each tier, Haiku, Sonnet, Opus, and the newer Mythos class, carries its own version number and moves on its own schedule. That is why there was never going to be a single launch day. The 5-generation crossed over one tier at a time across 2026, and the pattern repeated on 22 September when Claude Opus 5.5 opened a 5.5 generation with Sonnet and Haiku still to come.

Three moments matter. On June 9, 2026, Anthropic opened a fourth tier above Opus, the Mythos class, and made it publicly usable through Claude Fable 5. Three weeks later, on June 30, the Sonnet tier crossed into 5 with Claude Sonnet 5, covered at launch by TechCrunch. The Opus line held at 4.x for another month, moving 4.7 to 4.8 rather than jumping a whole number, until Claude Opus 5 finally landed on July 24 per Anthropic’s own announcement.

The Confirmed Claude Lineup (September 2026)

Here is every Claude model you can actually reach today, top tier first. The Mythos class still sits above the rest for long autonomous and security work, Opus 5.5 is the flagship and the highest-scoring Claude on the public index, Sonnet is the value-agent pick, and Haiku is the budget workhorse.

*Sonnet 5 launched on introductory pricing due to rise to $3/$15 on September 1, 2026; Anthropic cancelled that increase on August 10, 2026, so $2/$10 is permanent. Claude Opus 5 and Claude Opus 4.8 (May 28, 2026) both sit on Anthropic’s legacy list now, and both still answer at $5/$25. Full numbers are in our Claude pricing breakdown.

What Each “5” Model Actually Is

Claude Sonnet 5

Claude Sonnet 5 is the value pick of the group. Released June 30, 2026 with model ID claude-sonnet-5, it is built for agentic work, planning a sequence of steps, calling tools, reading the result, and continuing without a human nudging it at every turn. It ships with a 1M-token context window and a January 2026 knowledge cutoff, and it is the default model for Free and Pro users on Claude.ai, plus Claude Code, the API, Cursor, VS Code, and GitHub Copilot.

On agentic coding it scores around 63.2%, behind the Opus tier but well ahead of Sonnet 4.6, and on some knowledge-work tests it edged out Opus 4.8 outright at roughly a quarter of the price. For most people running agents or everyday coding, it is still the right default, and Opus 5.5 is the step up rather than the replacement. Anthropic has promised a Claude Sonnet 5.5 without naming a date.

Claude Fable 5 and Mythos 5 (the Mythos class)

Claude Fable 5 is the public face of Anthropic’s top tier, a fourth step that sits alongside Haiku, Sonnet, and Opus rather than inside the same ladder. It launched June 9, 2026 at $10/$50 per million tokens and is built for long-running autonomous work rather than quick chat. Its unrestricted sibling, Claude Mythos, keeps the same capability with the safety layer removed and stays gated to vetted partners.

Both models were briefly pulled offline on June 12 under a US export-control order, then restored after the controls were lifted on June 30. Fable 5 came back globally on July 1, while Mythos 5 returned only to a set of approved US organizations. The tier’s lead on the public boards has changed hands twice since: Opus 5 passed Fable 5 at half the price, the September 1, 2026 refresh to Fable 5.1 and Mythos 5.1 put the class back in front, and Opus 5.5 took the top spot back on September 22. What has not changed is offensive security work, where Mythos is still the stronger model, so the tier has not been made redundant.

Claude Opus 5, and the Opus 5.5 That Replaced It

Claude Opus 5 is the model that closed the generation. Released July 24, 2026 under the API id claude-opus-5, it held the same $5/$25 pricing as Opus 4.8 while landing at roughly half the cost of Fable 5, and it ran as the default model on Claude Max and the strongest model available on Claude Pro. That is no longer its job. Claude Opus 5.5 arrived on September 22, 2026 at $4/$20, took both of those slots, and Anthropic moved Opus 5 onto its legacy list the same day. The old model id still answers.

Two things were genuinely new when Opus 5 landed, and both carried forward into 5.5. A Fast mode runs at about 2.5 times the default speed for twice the base price, and an effort setting lets you dial from low up to a max tier, trading tokens against depth of reasoning. On Artificial Analysis Opus 5 took first place at launch on the v4.1 leaderboard, with 61 on the Intelligence Index. That index has been rescaled twice since, and on the current v4.3.2 board Opus 5 reads 50.8, behind Opus 5.5 on 57.6, Fable 5.1 on 53.4 and GPT-6 Astra on 52.7. Only ever compare scores that carry the same index version.

Anthropic’s own reported numbers point the same way, with more than double Opus 4.8’s score on Frontier-Bench v0.1, a CursorBench 3.2 result within 0.5% of Fable 5 at max effort, and roughly three times the next-best model on ARC-AGI 3. Those are vendor benchmarks, so treat them as directional. The one clear weak spot Anthropic discloses is security, where Opus 5 matches Mythos 5 at finding vulnerabilities but stays substantially behind it at exploiting them.

Where Claude 5 Goes From Here

With Opus at 5, every tier that carries a public version number had crossed into the generation, and the open question moved to what comes after it. Anthropic answered part of it on September 22 by opening the 5.5 family with Opus 5.5 and promising Claude Sonnet 5.5 and Claude Haiku 5.5 “in the coming weeks” with no date attached. Haiku is still the oldest model in the lineup at 4.5, dating to October 2025, so it has the most ground to make up.

The other open question, whether the Mythos tier would keep its lead, has now been answered twice. Opus 5 passed Fable 5 on the public leaderboards at half the price, Anthropic replied on September 1, 2026 with Fable 5.1 and Mythos 5.1, which put the class back in front, and then Opus 5.5 took the top of the Intelligence Index on September 22 at 57.6 against Fable 5.1’s 53.4, while costing $4/$20 against $10/$50. Treat any leaked model ID, price sheet, or benchmark for an unannounced Claude as fiction until it shows up in Anthropic’s own docs.

Which Claude Should You Use Right Now?

Picking well saves both money and time. Use Claude Opus 5.5 when the task is genuinely hard, deep debugging, subtle judgment calls, or research where a wrong answer is expensive, and turn the effort setting up when you would rather spend tokens than get it wrong. On Claude Pro and Claude Max it is already the strongest model you have.

Reach for Claude Sonnet 5 for agent loops, everyday coding, and anything that feeds a large context, since it delivers close to flagship quality at a fraction of the cost. Drop to Haiku 4.5 for high-volume, simple jobs like classification, log analysis, or chat triage, where paying Opus rates would just be burning money. Go to Fable 5.1 or Mythos 5.1 only for multi-day autonomous runs or serious security research, which is the one area where the Opus tier still does not lead.

And if you would rather not juggle four model choices or separate subscriptions, Fello AI puts every Claude tier alongside GPT, Gemini, Grok, and DeepSeek in one app for Mac, iPhone, and iPad, so you can switch models mid-task and always land on the right one.

Conclusion

“Claude 5” turned out to be a generation rather than a single model, and as of July 24, 2026 all of it is here. Sonnet 5, Fable 5, Mythos 5 and Claude Opus 5 are live, Fable 5.1 and Mythos 5.1 followed on September 1, 2026, and Claude Opus 5.5 opened the next generation on September 22. The practical move is to use Claude Sonnet 5 for value and agents and step up to Opus 5.5 for the hardest problems. To see how they stack up against everything else, check our guide to the best AI models right now, and if you want every tier in one place, start with Fello AI.

FAQ