# Which Claude Model? Opus 5.5 vs Sonnet 5.5 vs Haiku 5.5 vs Fable 5.2
> Opus 5.5 for most paid-plan work, Sonnet 5.5 for speed and Free, Haiku 5.5 for volume, Fable 5.2 for the hardest jobs. Prices and plan access, Nov 2026.
**Author:** [Alex Lowe](https://theaicareerlab.com/about) — Founder, The AI Career Lab
**Published:** 2026-05-20
**Updated:** 2026-11-13
**Last reviewed:** 2026-11-13
**Canonical URL:** https://theaicareerlab.com/blog/which-claude-model-should-you-use
**Category:** guide
**Tags:** Claude, Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, Claude Sonnet 5, Claude Haiku 4.5, Claude Fable 5.2, Claude Fable 5.1, model selection, 2026
---> **TL;DR.** On a paid plan, start with **Opus 5.5**. Anthropic's models page recommends it "for most workloads", and on claude.ai it costs nothing extra per message. **Sonnet 5.5** is the faster, half-price pick for quick tasks, the Free plan and cost-sensitive API work. **Haiku 5.5** is for speed and volume. **Fable 5.2** (released November 9, 2026, replacing Fable 5.1) is for the rare job where Opus 5.5 still falls short. **Sonnet 5** and **Haiku 4.5** are now legacy models: Sonnet 5.5 replaced Sonnet 5 on September 28, 2026 and Haiku 5.5 replaced Haiku 4.5 on October 7.

_Updated November 13, 2026 for Fable 5.2 (and October 19 for Sonnet 5.5 and Haiku 5.5). Prices, model IDs and specs are checked against Anthropic's [models overview](https://platform.claude.com/docs/en/about-claude/models/overview), [API pricing](https://platform.claude.com/docs/en/about-claude/pricing) and [claude.com/pricing](https://claude.com/pricing) on November 13, 2026._

| Use | Pick | Why |
|---|---|---|
| Most professional work on a paid plan: drafting, analysis, hard code, long agentic work | **Opus 5.5** | Anthropic: "start with Claude Opus 5.5 for most workloads"; built for long-running agentic coding and knowledge work; $4/$20, 1M context |
| Quick tasks, the Free plan, cost-sensitive API work | **Sonnet 5.5** | Anthropic's "best combination of speed and intelligence"; half Opus 5.5's price at $2/$10, faster, 1M context |
| Volume, speed, lowest cost | **Haiku 5.5** | Fastest Claude model; from $0.10/$0.50 (prompts up to 100K tokens), 1M context |
| The hardest, longest-running agent work when cost is secondary | **Fable 5.2** | Anthropic's most capable generally available model; $8/$40, 1M context |

Anthropic's current lineup, as of November 13, 2026, is **Fable 5.2**, **Opus 5.5**, **Sonnet 5.5** and **Haiku 5.5**. All four take text and images, use tools, have a 1M-token context window and 128K max output, and share a June 2026 knowledge cutoff. They differ on capability, speed and cost. If you last read this guide when it recommended Sonnet 5 or Haiku 4.5, skip to [what changed](#what-changed-sonnet-5-to-sonnet-55-and-haiku-45-to-haiku-55).

## The four current models at a glance

| | **Fable 5.2** | **Opus 5.5** | **Sonnet 5.5** | **Haiku 5.5** |
|---|---|---|---|---|
| **Position** | Top tier; most capable, for when cost is secondary | Long-running agentic coding and knowledge work | Best combination of speed and intelligence | Fastest; high-volume, latency-sensitive tasks |
| **Released** | Nov 9, 2026 | Sept 22, 2026 | Sept 28, 2026 | Oct 7, 2026 |
| **API ID** | `claude-fable-5-2` | `claude-opus-5-5` | `claude-sonnet-5-5` | `claude-haiku-5-5` |
| **Context window** | 1M tokens | 1M tokens | 1M tokens | 1M tokens |
| **Max output** | 128K tokens | 128K tokens | 128K tokens | 128K tokens |
| **Price (API, per M tokens)** | $8 input, $40 output | $4 input, $20 output | $2 input, $10 output | From $0.10 input, $0.50 output (prompts up to 100K tokens; $0.50/$2.50 above) |
| **Cache reads** | $0.20/M | $0.20/M | $0.10/M | $0.01/M |
| **Speed** | Moderate | Moderate | Fast | Fastest |
| **Thinking** | Adaptive, always on | Adaptive, always on | Adaptive | Adaptive |
| **Default effort (API)** | `high` | `medium` | `high` | `medium` |
| **Knowledge cutoff** | June 2026 | June 2026 | June 2026 | June 2026 |

## What changed: Sonnet 5 to Sonnet 5.5, and Haiku 4.5 to Haiku 5.5

Two of the four tiers turned over in the past month. Both older models still work on the API, but neither is the current pick.

| | **Sonnet 5** (legacy) | **Sonnet 5.5** (current) | **Haiku 4.5** (legacy) | **Haiku 5.5** (current) |
|---|---|---|---|---|
| **Released** | June 30, 2026 | Sept 28, 2026 | Oct 15, 2025 | Oct 7, 2026 |
| **Price (per M tokens)** | $2 / $10 | $2 / $10 | $1 / $5 | From $0.10 / $0.50 |
| **Cache reads** | $0.20/M | $0.10/M | $0.10/M | $0.01/M |
| **Context window** | 1M | 1M | 200K | 1M |
| **Max output** | 128K | 128K | 64K | 128K |
| **Thinking** | Adaptive | Adaptive | Manual extended thinking | Adaptive |
| **Knowledge cutoff** | Jan 2026 | June 2026 | Feb 2025 | June 2026 |
| **Earliest retirement** | June 30, 2027 | Sept 28, 2027 | Oct 15, 2026 | Oct 7, 2027 |

**Sonnet 5 → Sonnet 5.5.** Same list price, but Anthropic says Sonnet 5.5 runs more than 30% faster and costs up to 30% less per task, because it needs fewer tokens to do the same work. Cache reads are half the price. On Anthropic's own benchmarks the jump is large: GDPval-AA, a test of real-world work across 44 occupations, went from 1,449 to 1,844, against Opus 5.5's 1,846. Anthropic still says Opus 5.5 "remains clearly stronger at complex, open-ended work requiring sustained judgment." API users have breaking changes to handle (see the [Sonnet 5.5 migration guide](https://platform.claude.com/docs/en/models/sonnet-5-5/migration-guide)): forced tool use now returns an error, and code that turns thinking off must switch to the new `between_tools` setting. Our [Sonnet 5.5 write-up](/blog/claude-sonnet-5-5-for-professionals-2026) has the full comparison.

**Haiku 4.5 → Haiku 5.5.** A much bigger change. Haiku 5.5 lists at a tenth of Haiku 4.5's price for prompts up to 100K tokens, has five times the context window, twice the max output, adaptive thinking and a knowledge cutoff 16 months newer. Two catches: prompts over 100K tokens cost $0.50/$2.50, and Haiku 5.5 uses Anthropic's newer tokenizer, which counts the same text as roughly 30% more tokens. Compare real bills before and after, not just list prices. Read the [Haiku 5.5 migration guide](https://platform.claude.com/docs/en/models/haiku-5-5/migration-guide) before you swap the model string.

**What it means for the advice.** The default hasn't changed: Opus 5.5 on a paid plan. Where this guide used to say "Sonnet 5" for quick tasks and the Free plan, read **Sonnet 5.5**. Where it said "Haiku 4.5" for volume, read **Haiku 5.5**.

## Opus vs Sonnet vs Haiku: the tiers explained

Anthropic ships Claude in named tiers. The name tells you the trade-off between capability, cost and speed; the version number tells you which release you're on.

- **Opus** is the main high-end tier: the most capable of the three, built for hard reasoning, complex code and long agentic work where an early mistake compounds.
- **Sonnet** is the balanced middle tier: strong on most knowledge work at half Opus's price and faster.
- **Haiku** is the small tier: fastest and cheapest, for high-volume, real-time or background tasks where throughput matters more than depth.
- **Fable** sits above Opus. Anthropic calls Fable 5.2 (released November 9, 2026, replacing Fable 5.1) its most capable generally available model and prices it accordingly.

Price tracks the tiers: Haiku 5.5 from $0.10/$0.50, Sonnet 5.5 $2/$10, Opus 5.5 $4/$20, Fable 5.2 $8/$40 (Fable 5.1, now legacy, is $10/$50).

## The most important question: are you on claude.ai or the API?

**If you're on claude.ai** (the chat app), you pay per plan, not per token. On a paid plan (Pro, Max, Team or Enterprise), switch the picker to Opus 5.5 for most professional work: it's Anthropic's recommended starting point and costs you nothing extra per message. Switch to Sonnet 5.5 when you want a faster answer to a quick question. The Free plan includes Sonnet and Haiku but not Opus or Fable. Anthropic's Sonnet page says "anyone can chat with Claude using Sonnet 5.5 on Claude.ai", but Anthropic hasn't published which model each plan opens on, so check your picker. Pro is $20/month, or $17/month billed annually; see [Claude pricing](/blog/claude-pricing-plans-explained) for the other plans.

**If you're using the API** (or Claude Code on an API key), every request is billed by token. Anthropic's models page says that if you're unsure, start with Opus 5.5. At $4/$20 it costs twice Sonnet 5.5's $2/$10, so for high-volume or simple API work, test Sonnet 5.5 (or Haiku 5.5) on your own tasks before you commit.

If you want one default: **Opus 5.5 on a paid plan, Sonnet 5.5 on Free.** The rest of this guide covers the exceptions.

## Pick Opus 5.5 for: most professional work on a paid plan

Opus 5.5 is Anthropic's current Opus, released September 22, 2026. Anthropic describes it as built "for long-running agentic coding and knowledge work" and says it "performs at the level of Claude Fable 5.1 on most work." It clearly earns its place for:

- **Complex software engineering**: multi-file features, larger refactors, end-to-end feature work in Claude Code
- **Long multi-step agentic workflows**: compliance automation, data-extraction pipelines, anything that chains dozens of steps where one early error compounds
- **High-stakes deliverables**: when the document goes to a regulator, court or board and you want the strongest model your plan includes to draft or check it
- **Open-ended work that needs judgment**: the ambiguous brief, the long document where structure is the hard part, the analysis where you want the model to push back

Two notes for API users. **Fast mode** on Opus 5.5 costs $8/$40 per million tokens for faster output. The `effort` parameter **defaults to `medium`** on Opus 5.5; raise it for the hardest work.

### Profession-specific Opus 5.5 wins

- **Data scientists and technical operators** running hard multi-step pipelines in Claude Code
- **Healthcare compliance officers** running QSR gap audits across full DHF documents where a miss is costly
- **AI compliance officers** producing pre-legal regulatory screens across EU AI Act tiers, Annex III and US state rules
- **ESG sustainability analysts** mapping KPIs to several frameworks from a 200-page sustainability report

## Pick Sonnet 5.5 for: speed, the Free plan, and cost-sensitive work

Sonnet 5.5 is half Opus 5.5's API price, faster, and close behind it on many of Anthropic's benchmarks. Anthropic positions it as "a faster, lower-cost complement to Opus 5.5" that is "strongest at well-scoped everyday tasks, fixing bugs, and creating polished documents, slides, and spreadsheets." In practice:

- **Client, borrower and customer communication**: the daily back-and-forth, at speed
- **Well-defined documents**: memos, summaries, PRDs, slide decks built from a template
- **Long-document analysis**: full case files, policy documents, board packages, with the same 1M context as Opus
- **Iteration and polishing**: fast enough to think alongside you
- **API work at volume** where Haiku isn't deep enough and Opus costs too much

### Profession-specific Sonnet 5.5 wins

- **Loan officers** sending daily pipeline updates across 8+ active loans
- **Real estate agents** generating listing descriptions, client emails, CMAs and market updates
- **Attorneys and paralegals** drafting memos and summaries from source documents; check the hardest analyses on Opus 5.5
- **Copywriters** and **community managers** doing iteration-heavy, voice-sensitive work
- **Healthcare clinicians** drafting SOAP notes, treatment plans and patient education
- **Management consultants** turning meeting notes into strategy memos
- **AI product managers** structuring specs and rollout plans

## Pick Haiku 5.5 for: speed and scale

Haiku 5.5 (released October 7, 2026) is the fastest and cheapest current Claude model. Anthropic positions it "for high-volume, latency-sensitive tasks such as classification, extraction, and routing." It earns its place in:

- **High-volume classification or extraction**: thousands of inputs a day
- **Real-time chat**: in-product chatbots and support assistants where the reply has to feel instant
- **Background AI features**: the model running behind a feature, where latency is the user experience
- **Cost-sensitive workflows**: from $0.10/$0.50 for prompts up to 100K tokens, a twentieth of Sonnet 5.5's price

Haiku's constraints:

- **Long prompts cost more.** Above 100K tokens, input rises to $0.50/M and output to $2.50/M, still a quarter of Sonnet 5.5's rate.
- **New tokenizer.** The same text counts as roughly 30% more tokens than on Haiku 4.5.
- **Built for volume, not depth.** Keep quality-sensitive documents on Sonnet 5.5 or Opus 5.5.

### Profession-specific Haiku 5.5 wins

- **Customer support teams** building in-product AI chat where latency matters
- **Recruiters and HR** running high-volume resume classification or first-pass screening, with human review and EEOC-aware guardrails (see our [recruiter audit guide](/blog/ai-for-recruiters-guide))
- **Sales teams** generating high-volume personalized outreach from a strong template
- **Internal tooling**: Slack bots, internal knowledge search, the AI behind ops dashboards
- **Real-time content moderation** in community management

## Is there a Haiku 5?

**No. Anthropic went straight from Haiku 4.5 to Haiku 5.5.** Claude Haiku 5.5 (`claude-haiku-5-5`) came out on October 7, 2026. As of November 13, Anthropic's [models page](https://platform.claude.com/docs/en/about-claude/models/overview) lists it alongside Fable 5.2, Opus 5.5 and Sonnet 5.5, and Haiku 4.5 is a legacy model.

What that means for you:

- **For the cheapest, fastest model,** use **Haiku 5.5**: $0.10/M input and $0.50/M output for prompts up to 100K tokens ($0.50/$2.50 above), against Haiku 4.5's $1/$5.
- **Still on Haiku 4.5 in the API?** It keeps working for now. Anthropic's [deprecations page](https://platform.claude.com/docs/en/about-claude/model-deprecations) gave October 15, 2026 as the earliest possible retirement date. That date has passed and, as of November 13, Haiku 4.5 is still listed as active with no retirement date announced. Anthropic promises at least 60 days' notice before retiring a public model, so plan the move to Haiku 5.5 now rather than waiting for the email.

## Head-to-head: the matchups people search for

### Haiku 4.5 vs Sonnet 5.5

**Verdict: Sonnet 5.5 for quality; for speed and price, skip Haiku 4.5 and use Haiku 5.5.** Sonnet 5.5 has five times Haiku 4.5's context window (1M vs 200K), twice its max output (128K vs 64K), a knowledge cutoff 16 months newer (June 2026 vs February 2025) and adaptive thinking. Haiku 4.5's advantages were speed and price ($1/$5 against Sonnet 5.5's $2/$10), and Haiku 5.5 is faster still and costs about a tenth as much.

### Haiku 5.5 vs Sonnet 5.5

**Verdict: Sonnet 5.5 when a person reads and judges the output; Haiku 5.5 when the output feeds a pipeline, a chat widget or a loop that runs thousands of times a day.** Both have a 1M-token context window, 128K max output and a June 2026 cutoff. Sonnet 5.5 is the deeper model; Haiku 5.5 is faster and costs about a twentieth as much for prompts up to 100K tokens. On claude.ai this choice rarely matters: Haiku is a builder's model.

### Opus 5.5 vs Sonnet 5.5

**Verdict: Opus 5.5 by default on a paid plan; Sonnet 5.5 when speed, volume or API cost matters more.** Sonnet 5.5 comes close on several of Anthropic's benchmarks (GDPval-AA 1,844 vs 1,846; CursorBench 4.0 55.5% vs 57.8%) and beats Opus 5.5 on Terminal-Bench 4.0 (70.6% vs 66.4%). It trails more clearly on harder tests (FrontierCode 46.2% vs 54.4%; Humanity's Last Exam 64.5% vs 67.7%), and Anthropic says Opus 5.5 is "clearly stronger at complex, open-ended work requiring sustained judgment." Sonnet 5.5 is half the API price ($2/$10 vs $4/$20) and faster. On claude.ai there's no per-token cost, so stay on Opus 5.5 and drop to Sonnet 5.5 for quick tasks.

### Opus 5.5 vs Opus 5

**Verdict: Opus 5.5. It's cheaper and faster.** Opus 5.5 lists at $4/$20 against Opus 5's $5/$25, cache reads drop from $0.50 to $0.20 per million tokens, and Anthropic says typical workloads cost about 40% less. Output is more than 30% faster. On the API, Opus 5 stays available as a legacy model, but moving to `claude-opus-5-5` is not just a model-string swap: adaptive thinking is always on, forced tool use isn't supported, and the default effort is `medium`. Read [Anthropic's Opus 5.5 migration guide](https://platform.claude.com/docs/en/models/opus-5-5/migration-guide) first; our [Opus 5.5 write-up](/blog/claude-opus-5-5-for-professionals-2026) summarizes the changes.

### Fable 5.2 vs Opus 5.5

**Verdict: Opus 5.5 by default; Fable 5.2 for the most demanding reasoning and long-horizon agentic work.** That's Anthropic's own guidance: use Fable 5.2 "for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short." Fable 5.2 costs twice as much per token ($8/$40 vs $4/$20), but it scores higher on most of Anthropic's published benchmarks, and on one coding test Anthropic reports Fable 5.2 at high effort matched Opus 5.5 at max effort for roughly half the cost per task. On claude.ai, Opus 5.5 is included on every paid plan, while Fable 5.2 runs on usage credits on Pro and counts against a 50% weekly share on Max and Ultra. Our [Opus 5.5 vs Fable 5.2 guide](/blog/claude-opus-5-5-vs-fable-5-1-for-professionals) goes deeper.

### Fable 5.1 vs Opus 5.5 (legacy)

Fable 5.1 is now a legacy model. It costs $10/$50, 2.5 times Opus 5.5's price, and Anthropic said in September that Opus 5.5 "performs at the level of Claude Fable 5.1 on most work." If you're choosing between them today, compare Opus 5.5 with Fable 5.2 instead.

### Older matchups (legacy models)

These pairings still get searched. All of these models are still served on the API, but each has a newer, better option.

#### Sonnet 5 vs Haiku 4.5

The question from July to September 2026. Sonnet 5 was the better model for quality (1M context vs 200K, a January 2026 cutoff vs February 2025, adaptive thinking); Haiku 4.5 was the cheaper, faster one at half Sonnet 5's price ($1/$5 vs $2/$10). Both have been replaced: compare **Sonnet 5.5 vs Haiku 5.5** instead.

#### Haiku 5.5 vs Sonnet 5

A cross-generation pairing. Sonnet 5 is the deeper model, but it costs twenty times as much for prompts up to 100K tokens ($2/$10 vs $0.10/$0.50) and has an older knowledge cutoff (January 2026 vs June 2026). If you're on Sonnet 5, move to Sonnet 5.5: same price, faster, newer cutoff.

#### Haiku 5.5 vs Sonnet 4.6

Sonnet 4.6 (February 2026) is a legacy model at $3/$15 per million tokens, more than Sonnet 5.5's $2/$10. Use Haiku 5.5 for speed and price; if you need more depth, Sonnet 5.5 beats Sonnet 4.6 on price as well.

#### Opus 5 vs Sonnet 5

Opus 5 ($5/$25) led on hard code and long agentic chains; Sonnet 5 covered most knowledge work at less than half the price. Opus 5.5 and Sonnet 5.5 replaced both, so compare those two instead.

#### Opus 4.8 vs Sonnet 5

The question before Opus 5. Sonnet 5 matched Opus 4.8 on knowledge work, while Opus 4.8 led on complex software engineering (SWE-bench Pro: 69.2% vs 63.2%). The hard-code answer has since moved to Opus 5, then Opus 5.5.

#### Opus 4.8 vs Sonnet 4.6 and Haiku 4.5 vs Sonnet 4.6

All legacy now. Opus 5.5 costs less than Opus 4.8 did ($4/$20 vs $5/$25), Sonnet 5.5 costs less than Sonnet 4.6 ($2/$10 vs $3/$15), and Haiku 5.5 costs less than Haiku 4.5. The current choice is Opus 5.5, Sonnet 5.5 or Haiku 5.5.

## Where does Fable fit?

Fable 5.2 (`claude-fable-5-2`, released November 9, 2026) sits above the Opus tier. It's Anthropic's most capable generally available model, and it replaced Fable 5.1. Claude Mythos 5.2 is the same underlying model without Fable's extra cybersecurity and biology safeguards, offered only to organizations in Anthropic's verification programs. Fable 5.2 costs $8/$40 per million tokens with cache reads at $0.20 (fast mode $16/$80), and Anthropic says it generates output up to 30% faster than Fable 5.1. Fable 5.1 ($10/$50, cache reads $0.25) and Fable 5 ($10/$50, cache reads $1) are still available as legacy models, so there's no reason to pick either for new work.

The catch is billing. Fable isn't in Pro's plan limits: Pro users and Team standard seats pay for it with usage credits. Max and Ultra include Fable up to 50% of weekly limits (Team premium seats work the same way), and the Free plan has no access. The one exception was the launch week: from November 9 to November 16, 2026 (11:59 pm PT), Pro plans and Team standard seats got Fable inside their limits, up to 50% of weekly usage, plus one Fable reset.

Practical rule: treat Opus 5.5 as your ceiling for everyday work, since Anthropic recommends starting there for most coding and knowledge tasks, and save Fable 5.2 for sessions where the deepest reasoning or the longest agentic runs are worth paying for separately. See [What is Claude Fable 5.2?](/blog/what-is-claude-fable-5-2-2026) for the launch details; the earlier [Claude Fable 5.1 write-up](/blog/claude-fable-5-1-for-professionals-2026) and [Claude Fable 5 explainer](/blog/claude-fable-5-for-professionals) cover the previous versions, including the June export-control suspension.

## The decision tree

1. **Paid plan, professional work** (drafting, analysis, documents, hard code, agentic chains)? → **Opus 5.5**. Anthropic's recommended starting point.
2. **Free plan, or a quick task where speed matters more than depth** (or API work where cost does)? → **Sonnet 5.5**
3. **Is latency the user experience, or does this run thousands of times a day?** → **Haiku 5.5**
4. **Does Opus 5.5 at a higher effort still fall short, and does your plan (or budget) cover Fable?** → **Fable 5.2**

## When to switch mid-conversation

On claude.ai you can switch models mid-conversation, and the change applies from Claude's next reply. A pattern that works:

- **Work in Opus 5.5** for drafting, analysis and anything you'll put your name to
- **Drop to Sonnet 5.5** for quick lookups, reformatting and fast back-and-forth where you don't need the deepest answer
- **Step up to Fable 5.2** for the occasional session where maximum capability is worth it and your plan allows

## Effort levels: tuning depth within a model

Every current model takes an effort setting that controls how much it deliberates before answering. Lower effort is faster and uses less of your plan; higher effort suits long coding and agentic sessions. On claude.ai, click the model name next to the send button, then **Effort**; each model's recommended level is marked "Default" there. On the API the defaults are **`medium` on Opus 5.5 and Haiku 5.5** and **`high` on Sonnet 5.5 and Fable 5.2**. Anthropic says Sonnet 5.5 and Fable 5.2 default to medium in Claude Code and the Claude apps. [Anthropic's effort docs](https://platform.claude.com/docs/en/build-with-claude/effort) list the levels.

- **Pick the model by the type of task** (hard or high-stakes → Opus 5.5; quick or well-defined → Sonnet 5.5; volume → Haiku 5.5), **then use effort to tune** speed and cost within that model.
- **Before reaching for Fable 5.2,** try Opus 5.5 at a higher effort. That's Anthropic's own suggested order.
- On claude.ai, thinking can't be turned off for Haiku 5.5, Sonnet 5.5, Opus 5.5 or Fable 5.2. The legacy Haiku 4.5 is different: it uses manual extended thinking rather than effort-controlled adaptive thinking.

## What about the legacy models?

Anthropic still serves these older models on the API: **Fable 5.1**, **Sonnet 5**, **Haiku 4.5**, **Opus 5**, **Fable 5**, Opus 4.8, Opus 4.7, Opus 4.6, Opus 4.5 and Sonnet 4.6. A few dates to know, from Anthropic's [deprecations page](https://platform.claude.com/docs/en/about-claude/model-deprecations) as of November 13, 2026:

- **Sonnet 4.5** is deprecated and retires on **November 30, 2026**. Anthropic's recommended replacement is Sonnet 5.5.
- **Opus 4.5** can be retired from November 24, 2026 at the earliest; no date has been announced.
- **Haiku 4.5**'s earliest possible retirement date (October 15, 2026) has passed with no retirement announced.
- **Sonnet 4 and Opus 4** retired on June 15, 2026, and **Opus 4.1** retired on August 5, 2026.

If your integration pins `claude-sonnet-5` or `claude-sonnet-4-6`, move to `claude-sonnet-5-5` (same price as Sonnet 5, cheaper than Sonnet 4.6). If it pins `claude-haiku-4-5`, move to `claude-haiku-5-5`. If it pins `claude-opus-5` or `claude-opus-4-8`, Opus 5.5 is better and cheaper. If it pins `claude-fable-5-1`, Fable 5.2 is cheaper ($8/$40 vs $10/$50); re-run your evals before switching. Each of these moves has a migration guide; read it first.

## How API model versioning works

Model IDs are pinned snapshots, not pointers that move to the newest release. `claude-sonnet-5` won't turn into Sonnet 5.5 on its own, just as `claude-opus-5` stayed put when Opus 5.5 launched. For production that's what you want: your model doesn't change under you. It also means you have to update model strings yourself when a new version ships.

## Pricing context

Per-token API pricing as of November 13, 2026, from [Anthropic's pricing documentation](https://platform.claude.com/docs/en/about-claude/pricing):

- **Fable 5.2:** $8/M input, $40/M output; cache reads $0.20/M; Fast mode $16/$80
- **Opus 5.5:** $4/M input, $20/M output; cache reads $0.20/M; Fast mode $8/$40; Batch $2/$10. Anthropic says typical workloads cost about 40% less than on Opus 5
- **Sonnet 5.5:** $2/M input, $10/M output; cache reads $0.10/M; Batch $1/$5
- **Haiku 5.5:** $0.10/M input, $0.50/M output for prompts up to 100K tokens ($0.50/$2.50 above); cache reads $0.01/M
- **Legacy:** Fable 5.1 $10/$50 (cache reads $0.25); Sonnet 5 $2/$10 (cache reads $0.20); Haiku 4.5 $1/$5; Opus 5 $5/$25; Fable 5 $10/$50 (cache reads $1); Sonnet 4.6 $3/$15

If you use claude.ai rather than the API, per-token pricing mostly doesn't apply: you pay per plan and work within its usage limits. It matters if you run Claude Code on an API key or build Claude into your own product. [claude.com/pricing](https://claude.com/pricing) has the current plan details; check it before you commit to a plan.

## Bottom line

On a paid claude.ai plan, switch the picker to Opus 5.5: it's Anthropic's recommended starting point and costs nothing extra per message. Use Sonnet 5.5 on the Free plan or when speed matters more than depth, Haiku 5.5 when speed or volume is the point, and Fable 5.2 for the rare session where Opus 5.5 falls short. If you're still on Sonnet 5 or Haiku 4.5, the newer versions are faster, have newer knowledge, and cost the same or less.

Model is only half the decision. The other half is the *surface*: chat, Cowork, Claude Code or the Chrome extension. See [which Claude surface for which task](/blog/claude-chat-vs-cowork-vs-claude-code), or answer four questions in the [surface chooser](/tools/which-claude-surface).

For the deeper dives, see our [Claude Sonnet 5.5 write-up](/blog/claude-sonnet-5-5-for-professionals-2026), [Claude Opus 5.5 write-up](/blog/claude-opus-5-5-for-professionals-2026), [Opus 5.5 vs Fable 5.2 guide](/blog/claude-opus-5-5-vs-fable-5-1-for-professionals) and [Claude Fable 5.2 explainer](/blog/what-is-claude-fable-5-2-2026). The earlier [Claude Sonnet 5 launch write-up](/blog/claude-sonnet-5-for-professionals-2026) and [Claude Opus 5 write-up](/blog/claude-opus-5-for-professionals-2026) cover the previous generation. For Claude vs ChatGPT, see our [profession-specific comparison hub](/compare).

---

*Model specifications come from [Anthropic's models overview](https://platform.claude.com/docs/en/about-claude/models/overview) and per-model pages; API prices from [Anthropic's pricing documentation](https://platform.claude.com/docs/en/about-claude/pricing) and [claude.com/pricing](https://claude.com/pricing); lifecycle dates from the [model deprecations page](https://platform.claude.com/docs/en/about-claude/model-deprecations); Sonnet 5.5 launch claims and benchmarks from [Anthropic's Sonnet 5.5 announcement](https://www.anthropic.com/claude-sonnet-5-5); Opus 5.5 launch claims from [Anthropic's Opus 5.5 announcement](https://www.anthropic.com/claude-opus-5-5); and Fable 5.2 claims from [Anthropic's Fable 5.2 announcement](https://www.anthropic.com/claude-fable-and-mythos-5-2) and its [launch-offer help article](https://support.claude.com/en/articles/17506494). All checked November 13, 2026. Benchmark and cost-reduction figures are Anthropic's own published results. Anthropic changes models, prices and plans often, so verify current details before a purchase or integration decision.*
## Frequently asked questions

### Which Claude model should I use?

On a paid claude.ai plan, start with Opus 5.5. Anthropic's own models page says 'start with Claude Opus 5.5 for most workloads', and on a subscription there is no per-message cost. Use Sonnet 5.5 for quick tasks, on the Free plan, and for cost-sensitive API work. Use Haiku 5.5 for high-volume or real-time work. Use Fable 5.2 only when Opus 5.5 at a higher effort setting still falls short. As of November 13, 2026 those four are Anthropic's current models; Fable 5.1, Sonnet 5 and Haiku 4.5 are legacy.

### What is the newest Claude model?

As of November 13, 2026, Anthropic's models page lists four current models: Fable 5.2 (released November 9, 2026, the most capable and the newest), Opus 5.5 (September 22), Sonnet 5.5 (September 28) and Haiku 5.5 (October 7). Fable 5.2 replaced Fable 5.1, Sonnet 5.5 replaced Sonnet 5 and Haiku 5.5 replaced Haiku 4.5. The older models are still available on the API as legacy models.

### Haiku 4.5 vs Sonnet 5.5: which should I use?

For most people, neither question is the real choice anymore. Haiku 4.5 is now a legacy model, so the fast, cheap option is Haiku 5.5. If you are choosing between Haiku 4.5 and Sonnet 5.5 as they are: Sonnet 5.5 is far more capable, with a 1M-token context window (Haiku 4.5 has 200K), 128K max output (Haiku 4.5 has 64K) and a June 2026 knowledge cutoff (Haiku 4.5's is February 2025). Haiku 4.5 costs $1/$5 per million tokens against Sonnet 5.5's $2/$10. Haiku 5.5 costs less than either, from $0.10/$0.50, with a 1M context window and a June 2026 cutoff. Pick Sonnet 5.5 when quality matters and Haiku 5.5 when speed or volume does.

### Haiku 5.5 vs Sonnet 5: which is better?

They are from different generations. Haiku 5.5 (October 7, 2026) is the current small, fast model. Sonnet 5 (June 30, 2026) is now a legacy model, replaced by Sonnet 5.5 on September 28. Sonnet 5 is the deeper model of the two, but it costs $2/$10 per million tokens against Haiku 5.5's $0.10/$0.50 (prompts up to 100K tokens), and its knowledge cutoff is January 2026 against Haiku 5.5's June 2026. Both have a 1M-token context window. For new work, compare Haiku 5.5 with Sonnet 5.5 instead: Sonnet 5.5 costs the same as Sonnet 5 and is faster.

### Haiku 5.5 vs Sonnet 4.6: which should I use?

Haiku 5.5 for speed and price. Sonnet 4.6 is a legacy model from February 2026 that costs $3/$15 per million tokens, more than Sonnet 5.5's $2/$10. If you need more depth than Haiku 5.5, move to Sonnet 5.5, not Sonnet 4.6. Anthropic lists Sonnet 4.6's retirement as not sooner than February 17, 2027.

### Sonnet 5 vs Haiku 4.5: which should I use?

Both are now legacy models, so for new work compare Sonnet 5.5 with Haiku 5.5. Between the two older models: Sonnet 5 was the better choice for quality (1M context vs 200K, a January 2026 cutoff vs February 2025, adaptive thinking), and Haiku 4.5 was the cheaper, faster one at $1/$5 per million tokens against Sonnet 5's $2/$10. Both are still served on the API.

### Is Opus 5.5 better than Sonnet 5.5?

Yes, for complex and open-ended work. Anthropic says Opus 5.5 'remains clearly stronger at complex, open-ended work requiring sustained judgment', even though Sonnet 5.5 comes within a few points on several of its benchmarks. Sonnet 5.5 is half the API price ($2/$10 vs $4/$20) and faster. On a paid claude.ai plan, where there is no per-token cost, Opus 5.5 is the better default; Sonnet 5.5 is the pick for quick tasks, the Free plan and cost-sensitive API work.

### What changed between Sonnet 5 and Sonnet 5.5?

Sonnet 5.5 (released September 28, 2026) keeps Sonnet 5's $2/$10 price but, Anthropic says, runs more than 30% faster and costs up to 30% less per task because it uses fewer tokens. Cache reads dropped from $0.20 to $0.10 per million tokens, and the knowledge cutoff moved from January 2026 to June 2026. Context (1M) and max output (128K) are unchanged. API users should read the migration guide first: forced tool use now returns an error, and turning thinking off needs the new between_tools setting.

### Is there a Claude Haiku 5?

No. Anthropic skipped Haiku 5 and released Claude Haiku 5.5 (claude-haiku-5-5) on October 7, 2026. It costs from $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100K tokens ($0.50/$2.50 above that), has a 1M-token context window and a June 2026 knowledge cutoff. Haiku 4.5 is now a legacy model at $1/$5. Haiku 5.5 counts the same text as roughly 30% more tokens than Haiku 4.5, so compare real bills, not just list prices.

### Is Opus 5.5 better than Opus 5?

Yes, and it costs less. Opus 5.5 lists at $4/$20 per million tokens against Opus 5's $5/$25, generates output more than 30% faster, and Anthropic says it costs about 40% less on typical workloads. It is the current Opus on paid claude.ai plans. API users should read Anthropic's Opus 5.5 migration guide before switching: adaptive thinking is always on and forced tool use isn't supported.

### Which Claude model should I use on claude.ai?

On a paid plan (Pro, Max, Team or Enterprise), switch the model picker to Opus 5.5 for most work. There is no per-token cost on claude.ai, only per-plan usage limits, so you can switch back to Sonnet 5.5 for quick tasks. The Free plan includes Sonnet and Haiku but not Opus or Fable. Anthropic's Sonnet page says anyone can chat with Sonnet 5.5 on Claude.ai, but it hasn't published which model each plan opens on, so check your picker. Pro costs $20/month, or $17/month billed annually.

### How much does each Claude model cost?

API pricing as of November 13, 2026, per million input/output tokens: Fable 5.2 $8/$40 (cache reads $0.20; Fast mode $16/$80); Opus 5.5 $4/$20 (cache reads $0.20; Fast mode $8/$40); Sonnet 5.5 $2/$10 (cache reads $0.10); Haiku 5.5 $0.10/$0.50 for prompts up to 100K tokens and $0.50/$2.50 above (cache reads $0.01). Legacy: Fable 5.1 $10/$50, Sonnet 5 $2/$10, Haiku 4.5 $1/$5, Opus 5 $5/$25. On claude.ai you pay per plan, not per token: Free $0, Pro $20/month ($17 billed annually), Max from $100/month, Team and Enterprise per seat.

### What is the difference between Opus, Sonnet, and Haiku?

They are Anthropic's three main tiers. Opus is the most capable and most expensive of the three, for hard reasoning, complex code and long agentic work. Sonnet is the balanced middle tier: strong on most knowledge work, cheaper and faster. Haiku is the smallest, fastest and cheapest, for high-volume and real-time tasks. Today's versions are Opus 5.5, Sonnet 5.5 and Haiku 5.5, with Fable 5.2 as a fourth tier above Opus.

### What happened to Claude Fable 5, and can I use it now?

Yes, with caveats. Fable 5 launched June 9, 2026, was suspended June 12 under a US government export-control directive, and was restored globally on July 1. On September 1, 2026 Anthropic released Fable 5.1, which replaces it as the top tier at the same $10/$50 price with cheaper cache reads ($0.25 instead of $1). On November 9, 2026 Fable 5.2 replaced Fable 5.1 at a lower $8/$40; Fable 5.1 and Fable 5 stay available as legacy models. Fable isn't on the Free plan; Pro users pay for it with usage credits (apart from a launch week, November 9 to 16, 2026, when Fable was included), and Max and Ultra include it up to 50% of weekly limits. Anthropic still recommends starting with Opus 5.5 for most coding and knowledge work, and moving up to Fable 5.2 for the most ambitious, open-ended technical work.

---

*Canonical version: https://theaicareerlab.com/blog/which-claude-model-should-you-use*
*This document is the Markdown companion served for AI crawlers and answer engines. See the canonical URL for the rendered version with navigation, related content, and interactive elements.*