Model comparisons

Claude Opus 5.5 vs Fable 5.2 vs Sonnet 5.5: Which Should You Use?

Compare Claude Sonnet 5.5, Opus 5.5, and Fable 5.2 across Anthropic benchmarks, API pricing, effort defaults, plan access, Claude Code routing, and retention rules.

UpdatedNovember 9, 2026

In short

Start with Opus 5.5. Anthropic recommends it as the starting point for most coding and knowledge work, Claude Code uses it by default, and it costs $4 per million input tokens and $20 per million output tokens. Move fast, repetitive or high-volume work down to Sonnet 5.5, which costs half as much ($2 and $10). Move up to Fable 5.2 for the hardest open-ended technical work, or when Opus 5.5 at a higher effort setting still fails your own tests: it costs twice Opus 5.5's rates ($8 and $40), runs on usage credits on Claude Pro outside its launch week, and requires 30-day data retention on the API. All three have a 1M-token context window and 128K tokens of output, so size does not decide.

On this page4

Anthropic's figures put Fable 5.2 a step above Opus 5.5

Fable 5.2 launched on November 9, 2026, and Anthropic's launch table scores it above Opus 5.5 on almost every benchmark it reports. That reverses the picture at Opus 5.5's launch, when Anthropic said Opus 5.5 performed at the level of Fable 5.1. The rows below come from the Fable 5.2 table, with safeguards on and each model at its highest effort unless noted.

Benchmark (Anthropic's figures)	Fable 5.2	Opus 5.5	Fable 5.1
FrontierCode 2.0 (agentic coding)	49.1% (xhigh)	45.5%	32.4% (xhigh)
FrontierSWE v2 (agentic coding)	69.9%	62.3%	56.3%
AutomationBench (business workflows)	51.4%	42.5%	31.4%
OSWorld 2.1, offline set, partial (computer use)	89.3%	87.2%	83.5%
Humanity's Last Exam, with tools	74.2%	67.7%	65.6%
Remote Labor Index (knowledge work)	41.3% (high)	21.3%	17.9%
AA-Briefcase v1.1 (knowledge work, Elo)	1785	1800	1669

Check prices on claude.com

Fable 5.2's largest leads are on long, multi-step work: the Remote Labor Index score nearly doubles Opus 5.5's, and AutomationBench and FrontierSWE move by 7 to 9 points. On computer use the gap is about two points, and on AA-Briefcase Opus 5.5 is ahead. The Fable 5.2 table has no Sonnet 5.5 column; in Sonnet 5.5's own launch table it beats Opus 5.5 only on Terminal-Bench 4.0 (70.6 against 66.4), which Claude Sonnet 5.5 vs Opus 5.5 covers.

The cost side matters as much as the scores. On FrontierCode 2.0, Anthropic reports Fable 5.2 at high effort scoring 45.3%, about level with Opus 5.5's best 45.5% at max, for roughly half the cost per task. For hard coding tasks you already run on Opus 5.5 at max, that makes Fable 5.2 at high the comparison to test, not only Fable 5.2 at its top setting.

Fable 5.2 costs twice Opus 5.5 per token, with the same cache-read price

Prices are per million tokens on Anthropic's API. Read across a row: Opus 5.5 doubles Sonnet 5.5 and Fable 5.2 doubles Opus 5.5, except for cache reads, where Fable 5.2 and Opus 5.5 both charge $0.20.

	Sonnet 5.5	Opus 5.5	Fable 5.2
Input	$2	$4	$8
Output	$10	$20	$40
Cache read	$0.10	$0.20	$0.20
Batch input / output	$1 / $5	$2 / $10	$4 / $20
Fast mode input / output	Not offered	$8 / $40	$16 / $80

Fable 5.2 is 20% cheaper than Fable 5.1, which stays available at $10 and $50 as a legacy model. List prices are still not costs per task. Thinking tokens are billed as output, and the default effort differs: on the API, Opus 5.5 defaults to medium while Sonnet 5.5 and Fable 5.2 default to high. In Claude apps and Claude Code, Fable 5.2 runs at medium by default. Set effort explicitly and compare the cost of finished tasks.

Fast mode gives up to 2.5 times the output speed at twice the standard rate. For Fable 5.2 it is available in Claude Code, on Anthropic's API, on Google Cloud and on Microsoft Foundry. For Opus 5.5 it is a research preview on Anthropic's own API that needs access approval. In Claude Code on a subscription, fast mode runs on usage credits only.

Fable 5.2 is outside the plan on Pro and never the Claude Code default

The table shows which Claude plans include each model family, and where Fable runs on usage credits instead of the plan's limits. Fable 5.2, 5.1 and 5 follow the same rules.

	Sonnet	Opus	Fable
Free	Yes	No	No
Pro ($20 a month)	Yes	Yes	Usage credits (inside the plan November 9–16, 2026)
Max ($100) and Ultra ($200)	Yes	Yes	Up to 50% of weekly limits
Team	Yes	Yes	Premium seats; usage credits on Standard seats
Claude Code	`/model sonnet`	Default	`/model fable` (v2.1.321 or later)

On Max, Ultra and Team Premium seats, Fable use counts against the normal weekly limit, up to half of it, and draws it down faster than other models; it is a cap, not extra usage. For the launch, those plans also get one to three Fable limit resets, depending on how long the account has paid, which expire on March 8, 2027. Claude Code uses Opus 5.5 by default on Pro, Max, Team, Enterprise and the API, and no Fable model is the default on any plan or provider. On Enterprise, Fable 5.2 stays off until an admin enables it. Requests that Fable's safety classifiers block are answered by another model (Opus 4.8 for cybersecurity, Opus 5 for biology); Anthropic says Fable 5.2 blocked about 55% fewer requests than Fable 5.1 in early testing. On Amazon Bedrock and Google Cloud, Claude Code's sonnet alias still points to Sonnet 4.5, so set the full model ID there. Those clouds also set their own prices.

Data handling differs too. On the API, Fable 5.2 requires 30-day retention, and zero data retention is not available for it unless Anthropic expressly authorizes it. Opus 5.5 and Sonnet 5.5 prompts and outputs are not retained by default.

Checks before moving work off Opus 5.5

Up to Fable 5.2: re-run the failing tasks on Opus 5.5 at xhigh or max, then on Fable 5.2 at high, and compare the cost of finished tasks rather than list prices. Confirm that 30-day retention is acceptable and, on Pro or a Team Standard seat, that you will pay usage credits. Integrations built on Fable 5.1 need code changes as well as a new model ID; Claude Fable 5.2 vs Fable 5.1 lists them.
Down to Sonnet 5.5: set effort explicitly. If you want it to act without thinking first, between_tools turns off up-front thinking at high effort or below.
Further down: for classification, extraction and routing Anthropic names Haiku 5.5 instead; see Claude Haiku 5.5 vs GPT-6 Luna.

Claude usage limits by plan explains the five-hour and weekly limits these plans share.

Edited byRick Finch

Official sources

Editorial guidance grounded in official product sources.

Show all 24 sources

FAQ

Common questions

Which Claude 5 model should most people start with?

Start with Opus 5.5 ($4/$20 per million tokens): Anthropic recommends it as the starting point for most coding and knowledge work, and Claude Code uses it by default. Move fast, repetitive or high-volume work down to Sonnet 5.5 ($2/$10). Move up to Fable 5.2 ($8/$40) for the most ambitious open-ended technical work, or when Opus 5.5 at a higher effort setting still fails your own tests.

Is Fable 5.2 worth twice Opus 5.5's API price?

Sometimes, and per finished task rather than per token. Fable 5.2 lists twice Opus 5.5's input and output rates ($8/$40 against $4/$20), with the same $0.20 cache reads. On Anthropic's FrontierCode 2.0 test, Fable 5.2 at high effort scored about level with Opus 5.5 at max effort for roughly half the cost per task. Those are Anthropic's figures; repeat the comparison on your own tasks before moving work.

Do Fable 5.2, Opus 5.5, and Sonnet 5.5 have different context limits?

No. Anthropic lists a 1-million-token context window and a 128,000-token maximum output for all three. Context size does not decide among them; capability, effort, price, plan access, retention and safeguards do.

Can I use Sonnet 5.5, Opus 5.5, and Fable 5.2 in Claude Code?

Yes, subject to the account and provider. Opus 5.5 is the Claude Code default. Sonnet 5.5 needs Claude Code 2.1.284 or later and is what the sonnet alias selects on the Anthropic API. Fable 5.2 needs Claude Code 2.1.321 or later, the fable alias selects it, and it is not the default on any plan. On Pro and Team standard seats it bills usage credits outside the launch week that ends November 16, 2026.

Can eligible ZDR customers use Fable 5.2?

Usually not yet. Fable 5.2 requires 30-day data retention, and zero data retention is unavailable unless Anthropic expressly authorizes it. Enterprise Frontier Safeguards, in private beta, runs Anthropic's safety monitoring inside the customer's infrastructure so that eligible customers can keep zero data retention; ask the account team before assuming your agreement qualifies.

Should Fable 5.1, Mythos 5.2 or Opus 4.8 still be part of this choice?

Usually not. Fable 5.1 remains available as a legacy model at $10/$50, more than Fable 5.2 costs. Mythos 5.2 is the same model as Fable 5.2 at the same price with fewer cybersecurity and biology safeguards, and Anthropic offers it only to organizations verified through its Cyber and Life Sciences Verification Programs. Opus 4.8 is a legacy model that costs $5/$25, more than Opus 5.5.

Next steps