The latest Claude model is Claude Haiku 5.5, released on October 7, 2026. The current Claude lineup is Opus 5.5 (September 22, 2026), Sonnet 5.5 (September 28), Haiku 5.5 (October 7) and Fable 5.1 (September 1), the public version of Anthropic’s top Mythos tier. For most demanding work, Claude Opus 5.5 is the model to use: it performs close to Fable 5.1 on most tasks at less than half the API price.
Anthropic released nine new Claude models between June and October 2026, so it’s easy to lose track. This page lists every current model, what each one costs, what it’s best at, and which one fits your work. We update it when Anthropic releases a new model.
How we checked this
- Release dates, model IDs and prices come from Anthropic’s Claude Platform release notes, checked on November 2, 2026, plus Anthropic’s launch announcements as reported by the press.
- Performance comparisons use Anthropic’s own statements and independent benchmark results from Artificial Analysis, as reported in The Batch.
- No new Claude model had been released after October 7, 2026, when this page was last updated.
Current Claude Models at a Glance
| Model | Released | API model ID | API price per 1M tokens (input / output) | Best for |
|---|---|---|---|---|
| Claude Haiku 5.5 | October 7, 2026 | claude-haiku-5-5 |
About $0.10 / $0.50 | Fast, cheap, high-volume tasks |
| Claude Sonnet 5.5 | September 28, 2026 | claude-sonnet-5-5 |
$2 / $10 | Everyday writing, coding, documents and agents |
| Claude Opus 5.5 | September 22, 2026 | claude-opus-5-5 |
$4 / $20 | Complex reasoning, coding and long projects |
| Claude Fable 5.1 | September 1, 2026 | claude-fable-5-1 |
$10 / $50 | The hardest tasks; public Mythos-class model |
| Claude Mythos 5.1 | September 1, 2026 | claude-mythos-5-1 |
$10 / $50 | Vetted partners only (cybersecurity, life sciences) |
All current models handle text and images, and the 5.5 models accept up to 1 million input tokens and return up to 128,000 output tokens. Batch requests cost half price.
Claude Opus 5.5: The Best Model for Most Demanding Work
Claude Opus 5.5 launched on September 22, 2026, replacing Claude Opus 5. According to Anthropic, it is its best-performing model, performs similarly to Fable 5.1 on most work, and is particularly strong at finding problems in software. It also needs less computing power to run, which is why Anthropic cut its price by 20% compared with Opus 5.
- API price: $4 per million input tokens and $20 per million output tokens (Opus 5 was $5 and $25). Cache reads cost $0.20 per million.
- Context: 1 million tokens in, up to 128,000 out.
- In the Claude app: Anthropic raised the five-hour usage limits on Pro, Max and Team plans with this launch.
- What changed for users: Anthropic says Opus 5.5 communicates more clearly than Opus 5, in response to feedback that the earlier model’s answers were harder to follow.
Use it for: complex analysis, long documents, difficult coding, multi-step agent work, and anything where a wrong answer is expensive.
Claude Sonnet 5.5: The Everyday Workhorse
Claude Sonnet 5.5 arrived on September 28, 2026, replacing Claude Sonnet 5.
- API price: $2 per million input tokens and $10 per million output tokens. Cache reads dropped to $0.10 per million on October 7.
- Context: 1 million tokens in, up to 128,000 out.
- Strengths: Anthropic positions it for everyday development and office work such as documents, slides and spreadsheets. In independent tests reported by The Batch, Sonnet 5.5 at its highest reasoning setting ranked first on Artificial Analysis’s Coding Agent Index and second on its Intelligence Index, behind only Opus 5.5.
- Trade-off: at maximum reasoning it can use a lot of output tokens per task, so set a lower effort level for routine jobs.
Use it for: writing, email, research summaries, routine coding and most business tasks, at half the price of Opus.
Claude Haiku 5.5: The Newest and Cheapest
Claude Haiku 5.5 is the most recent Claude release, launched on October 7, 2026, and it replaces Claude Haiku 4.5.
- API price: about $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, with a higher rate for longer prompts. That’s around 75% cheaper than Haiku 4.5 on average.
- Context: 1 million tokens in, up to 128,000 out.
- Speed: Artificial Analysis measured it as the fastest model it had tested in tokens per second.
Use it for: customer support replies, sorting and tagging data, summarizing large batches of text, and running many small AI tasks inside an automation.
Claude Fable 5.1 and the Mythos Tier
Above Opus sits Anthropic’s Mythos tier. Claude Fable 5.1 and Claude Mythos 5.1 were released on September 1, 2026, replacing Fable 5 and Mythos 5 from June.
- They are the same underlying model. Fable 5.1 is the generally available version, with extra safeguards for sensitive areas such as cybersecurity and biology. Mythos 5.1 has fewer of those restrictions and is offered only to vetted organizations through trusted-access programs.
- Price: $10 per million input tokens and $50 per million output tokens.
- In the Claude app: Fable 5.1 is available on paid plans (Pro, Max, Team and Enterprise), not on the free plan.
- Project Glasswing: Anthropic’s program that gives a small number of trusted organizations access to Mythos-class models, including Claude Mythos Preview, for defensive cybersecurity and similar work. Anthropic’s Glasswing page has details.
Shortly after the first Mythos-class launch, Anthropic suspended access to Fable 5 and Mythos 5 on June 12, 2026, to comply with US Department of Commerce export controls. The controls were lifted on June 30, and Anthropic restored access on July 1.
Should you use Fable 5.1? For most people, no. Opus 5.5 performs similarly on most work and costs 60% less. In independent Artificial Analysis testing, Opus 5.5 even scored higher on the Intelligence Index (58 vs 53). Fable 5.1 is worth testing for the very hardest problems, where you can compare both on your own tasks.
Claude Model Release Timeline (2026)
| Date | Model | Tier |
|---|---|---|
| February 5 | Claude Opus 4.6 | Opus |
| February 17 | Claude Sonnet 4.6 | Sonnet |
| April 16 | Claude Opus 4.7 | Opus |
| May 28 | Claude Opus 4.8 | Opus |
| June 9 | Claude Fable 5 and Mythos 5 | Mythos (new) |
| June 30 | Claude Sonnet 5 | Sonnet |
| July 24 | Claude Opus 5 | Opus |
| September 1 | Claude Fable 5.1 and Mythos 5.1 | Mythos |
| September 22 | Claude Opus 5.5 | Opus |
| September 28 | Claude Sonnet 5.5 | Sonnet |
| October 7 | Claude Haiku 5.5 | Haiku |
Sources: Anthropic’s release notes and the Anthropic model release list at Opper.
Which Claude Model Should You Use?
| If you want to… | Use |
|---|---|
| Write emails, posts and proposals, or summarize documents | Sonnet 5.5 |
| Analyze a long contract, report or dataset carefully | Opus 5.5 |
| Build or debug complex code, or run long multi-step agents | Opus 5.5 |
| Answer customer messages or tag data at high volume | Haiku 5.5 |
| Run many small AI steps inside Zapier, Make or n8n | Haiku 5.5 |
| Tackle the hardest research or engineering problems | Opus 5.5, then test Fable 5.1 |
| Keep API costs as low as possible | Haiku 5.5 |
For freelancers and small businesses using the Claude app: Sonnet 5.5 handles most daily work. Switch to Opus 5.5 for important documents, tricky analysis or anything you’d otherwise double-check for an hour.
What Each Model Costs for a Real Task
Prices per million tokens are hard to picture. Here’s what 1,000 customer support replies would cost through the API, assuming each reply reads about 2,000 tokens (the customer’s message, history and your help docs) and writes about 500 tokens.
| Model | Input cost (2M tokens) | Output cost (0.5M tokens) | Total for 1,000 replies |
|---|---|---|---|
| Haiku 5.5 | $0.20 | $0.25 | About $0.45 |
| Sonnet 5.5 | $4.00 | $5.00 | $9.00 |
| Opus 5.5 | $8.00 | $10.00 | $18.00 |
| Fable 5.1 | $20.00 | $25.00 | $45.00 |
Batch processing halves these numbers, and reusing the same help docs with prompt caching cuts the input cost further. To estimate your own workload, try our AI API cost calculator.
How to Choose Your Claude Model
In the Claude app (web, desktop and mobile), open the model picker next to the message box and choose the model. Paid plans show more models and higher usage limits than the free plan.
In the API, set the model field to the model ID, for example claude-sonnet-5-5. Current models, including Sonnet 5.5 and Haiku 5.5, support adjustable reasoning effort with five levels, from low to max. Lower effort is faster and cheaper; higher effort is better for hard problems.
Check retirement dates. Anthropic retires older models on a schedule. Claude Sonnet 4.5 is due to be retired from the API on November 30, 2026, and Claude Opus 4.1, Sonnet 4 and Opus 4 have already been retired. If an app or automation of yours still uses an older model ID, switch it before that date.
How Claude Compares With Other AI Models
The October 2026 releases put Claude near the top of most independent rankings. In the tests reported by The Batch, Claude models led Artificial Analysis’s Coding Agent and Intelligence indexes, while Google’s Gemini 4 Argon narrowly led the Vals Index. Rankings change with every release, so choose based on your own tasks rather than one leaderboard.
If you’re deciding between assistants rather than models, our comparisons of ChatGPT vs Claude for cold emails and Perplexity vs Gemini test them on real work, and our guide on how to choose the right AI tool walks through the decision. Before sharing client data with any model, read whether it’s safe to put client information into an AI chatbot.
Frequently Asked Questions
What is the latest Claude model?
As of November 2026, the newest Claude model is Claude Haiku 5.5, released on October 7, 2026. The current lineup is Claude Opus 5.5 (September 22), Claude Sonnet 5.5 (September 28), Claude Haiku 5.5 (October 7) and Claude Fable 5.1 (September 1), the public Mythos-class model. Opus 5.5 is the strongest choice for most demanding work.
Which Claude model is the most powerful?
Anthropic’s most capable tier is Mythos. Claude Fable 5.1 is the public Mythos-class model, while Claude Mythos 5.1 is limited to vetted partners. In practice, Claude Opus 5.5 performs similarly to Fable 5.1 on most work and costs less than half as much through the API, so it is the best pick for most people.
How much does the Claude API cost?
Per million tokens, Claude Haiku 5.5 costs about $0.10 input and $0.50 output, Sonnet 5.5 costs $2 and $10, Opus 5.5 costs $4 and $20, and Fable 5.1 costs $10 and $50. Batch requests are half price, and cached input is much cheaper.
What is the difference between Claude Opus, Sonnet and Haiku?
Opus is Anthropic’s most capable general model for complex reasoning, coding and long tasks. Sonnet balances quality, speed and price for everyday work. Haiku is the smallest, fastest and cheapest, built for high-volume jobs like summaries, classification and customer support.
What are Claude Fable and Claude Mythos?
Mythos is a tier above Opus. Claude Fable 5.1 and Claude Mythos 5.1 are the same underlying model; Fable 5.1 is the generally available version with extra safeguards for areas like cybersecurity and biology, while Mythos 5.1 has fewer of those restrictions and is only available to vetted partners through programs such as Project Glasswing.
Is the latest Claude model free?
The Claude app has a free plan, but the most capable models and higher usage limits come with paid plans such as Pro, Max, Team and Enterprise. Claude Fable 5.1 is not available on the free plan. Developers pay per token through the API.
Which Claude models are being retired?
Anthropic retires older models on a schedule. Claude Sonnet 4.5 is set to be retired from the Claude API on November 30, 2026, and Claude Opus 4.1, Sonnet 4 and Opus 4 have already been retired. If you build on the API, move to the 5.5 models before a retirement date.
The Bottom Line
The newest Claude model is Haiku 5.5, but the one most people should use for serious work is Claude Opus 5.5, with Sonnet 5.5 for everyday tasks and Haiku 5.5 for high-volume automation. Bookmark this page: we update it whenever Anthropic releases a new model.
