Official sources · checked 2026-11-12

Claude Fable 5.2 vs Claude Opus 5.5 (2026-11)

As of 2026-11-12, Anthropic positions Fable 5.2 as its most capable model open to all customers, and Opus 5.5 as the model for long-running agentic coding and knowledge work; its models overview suggests starting most workloads on Opus 5.5. Both take 1M tokens of context and return up to 128K tokens, at official prices of $8 / $40 and $4 / $20 per million tokens, with default effort at high and medium. On QCode both run on the same key, and /model switches between them in Claude Code. Below is a point-by-point comparison from the official pages, plus a selection guide derived from those official statements.

Updated 2026-11-13

Switch models on the same endpoint and measure the cost per token yourself; check /models for callable models.

Billed per token — live rates on /models
Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
Self-service: top up and activate instantly

Create account View pricing

#Same specs: 1M / 128K#2026-11-09 / 2026-09-22#Default effort: high / medium#One key, switch anytime

Four official points

$8 / $40

Fable 5.2 official price (per million tokens)

Anthropic's pricing page and the Fable 5.2 overview both list $8 / MTok input, $40 / MTok output and $0.20 / MTok cache reads; the Batch API takes a 50% discount on input and output. These are official prices, checked 2026-11-12.

$4 / $20

Opus 5.5 official price (per million tokens)

Anthropic's pricing page and the Opus 5.5 overview both list $4 / MTok input, $20 / MTok output and $0.20 / MTok cache reads; the Batch API takes the same 50% discount on input and output. Price details for both models are on /claude-fable-5-2-pricing.

1M / 128K

Same context window and max output

The All models table in the official models overview gives both models 1M tokens of context and 128K tokens of max output, with a reliable knowledge cutoff of June 2026; QCode's /models also shows claude-fable-5-2 and claude-opus-5-5 as 1M / 128K. The one documented output difference is the 300K-output beta for the Batch API: the official list includes Opus 5.5 and does not include Fable 5.2.

49.1% / 45.5%

Vendor-reported FrontierCode 2.0

Anthropic's launch announcement benchmark table (vendor-reported, checked 2026-11-12) gives Fable 5.2 49.1% on FrontierCode 2.0 (the table labels this figure Xhigh effort) and Opus 5.5 45.5% (no effort label); the same table gives 69.9% and 62.3% on FrontierSWE v2, and 51.4% and 42.5% on AutomationBench. All of these come from one vendor's single table, and the announcement says Fable 5.2 is a step above Opus 5.5 in almost all the capabilities evaluated.

Where the official positioning differs

The official models overview separates them in two lines: Fable 5.2 is for demanding reasoning and long-horizon agentic work, while Opus 5.5 is for long-running agentic coding and knowledge work. The overview pages list Fable 5.2's typical uses as agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet or deck; Opus 5.5's typical uses are multihour autonomous coding agents, large-scale refactoring, complex systems engineering, vision-heavy workflows and computer use. The model-choice guide adds that Fable 5.2 is the most capable model open to all customers.

Shared API behavior, and four differences when moving from Opus 5.5 to Fable 5.2

The official migration guide says both models use the same Messages API and tool-use patterns, both always think adaptively, and both return a 400 error for forced tool choice. Moving traffic from Opus 5.5 to Fable 5.2 brings four differences: ① Fable 5.2 requires 30-day data retention and is not available under a zero data retention arrangement unless Anthropic expressly authorizes it; ② the prefix check applies to every Fable 5.2 account, while Opus 5.5 enforces it by default only for accounts created on or after 2026-08-31; ③ moving from Fable 5.2 back to Opus 5.5 means Opus cannot read Fable's thinking blocks, so the API drops them, the request still succeeds and the dropped blocks are not billed, but cost and latency can rise on the first turn after the switch, while in the other direction Fable 5.2 can read Opus 5.5's thinking blocks; ④ Fable 5.2's server-side fallback targets are only Opus 4.8 and Opus 5. The migration checklist also says to check your rate limits before moving production traffic: on Anthropic's API, Fable 5.2's rate limit is separate from Opus 5.5's, with lower token limits at each tier.

Timeline

2026-09-22

Anthropic released Claude Opus 5.5 with model ID claude-opus-5-5, positioned for long-running agentic coding and knowledge work.

2026-11-09

Anthropic released Claude Fable 5.2 with model ID claude-fable-5-2, positioned for demanding reasoning and long-horizon agentic work; the launch announcement also says most coding and knowledge tasks should still start on Opus 5.5.

2026-11-12

QCode's /models lists both claude-fable-5-2 and claude-opus-5-5, each shown as 1M / 128K; the official statements and QCode specs on this page were checked that day.

What the official pages confirm, and what you still have to test

Confirmed (official text)

① Release: Fable 5.2 shipped on 2026-11-09 with model ID claude-fable-5-2; Opus 5.5 shipped on 2026-09-22 with model ID claude-opus-5-5. ② Positioning: Fable 5.2 is for demanding reasoning and long-horizon agentic work and is the most capable model open to all customers; Opus 5.5 is for long-running agentic coding and knowledge work. ③ Specs: both have 1M tokens of context, 128K tokens of max output, a comparative latency of Moderate, and a reliable knowledge cutoff of June 2026. ④ Official prices per million tokens, checked 2026-11-12: Fable 5.2 is $8 input, $40 output and $0.20 cache reads; Opus 5.5 is $4 input, $20 output and $0.20 cache reads; the Batch API takes a 50% discount on input and output on both. ⑤ Thinking: both always think adaptively and cannot turn thinking off, Fable 5.2 defaults to high effort and supports all five effort levels, while Opus 5.5 defaults to medium effort. ⑥ Lifecycle: neither retires before 2027, with Fable 5.2 not sooner than 2027-11-09 and Opus 5.5 not sooner than 2027-09-22. ⑦ Benchmarks: in the launch announcement's single table, FrontierCode 2.0 is 49.1% vs 45.5%, FrontierSWE v2 is 69.9% vs 62.3%, and AutomationBench is 51.4% vs 42.5%, all vendor-reported. ⑧ Anthropic's official pages describe Claude Mythos 5.2 as offering the same capabilities as Fable 5.2 only to organizations verified through its programs such as the Cyber Verification Program; this page covers only claude-fable-5-2, which is open to all customers.

What you still have to verify beyond the benchmarks

The benchmark figures in the announcement are vendor-reported and all come from one vendor's single table, with the announcement's own qualifier being almost all the capabilities we evaluated. The two models also have different default effort levels (high and medium), and the effort labels on benchmark rows are not identical (on FrontierCode 2.0 the table labels Fable 5.2 as Xhigh effort and gives Opus 5.5 no label), so the numbers are not a same-setting comparison. That is why the official model-choice guide says to test on your own evals before moving from Opus 5.5 to Fable 5.2: one step up on a vendor benchmark does not mean one step up on your workload.

How to choose (from the official statements)

You need the highest available capability and long-horizon reasoning

The official model-selection matrix maps the highest available capability to Fable 5.2, with examples such as agent sessions that run for hours, multistep deep research, and analysis carried through to a finished deliverable; the models overview also says to move to Fable 5.2 when your evals on Opus 5.5 at higher effort still fall short. If you need the highest available capability and long-horizon reasoning, choose Fable 5.2.

Complex agentic coding, enterprise work, computer use

The official model-selection matrix maps complex agentic coding and enterprise work to Opus 5.5, with examples such as multihour autonomous coding agents, large-scale refactoring, complex systems engineering, vision-heavy workflows and computer use; both the models overview and the launch announcement suggest starting most workloads on Opus 5.5. If you need those, choose Opus 5.5. For the Sonnet 5.5 vs Opus 5.5 comparison, see /claude-sonnet-5-5-vs-opus-5-5.

Switching back and forth in Claude Code

Claude Code's official docs say: during a session, use /model with an alias or model name to switch immediately, or run /model with no argument to open the picker. With ANTHROPIC_BASE_URL pointed at QCode, select the full model names: /model claude-fable-5-2 for Fable 5.2 and /model claude-opus-5-5 for Opus 5.5. The official aliases for a direct Anthropic connection are /model fable (Fable 5.2 everywhere except Claude apps gateway sessions) and /model opus (Opus 5.5 on the Anthropic API). To save the choice as the default for new sessions, /model writes the model field in your user settings; --model and ANTHROPIC_MODEL apply only to the session you launch with them. ANTHROPIC_BASE_URL changes where requests are sent, not which model answers them. For the full Claude Code setup on QCode, see /claude-code-fable-5-2-setup.

The two models on QCode

QCode supports claude-fable-5-2 and claude-opus-5-5: in Claude Code and other Anthropic-protocol tools, set the model to either ID with Base URL https://api.qcode.cc/api; the same key also calls claude-fable-5-1, claude-fable-5 and more. One key covers both Anthropic-compatible and OpenAI-compatible endpoints (Claude models use the Anthropic endpoint above). Billing is per token, with per-model rates on /models. Not ready for a monthly plan? Pay-as-you-go Packs start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.

Frequently asked questions

Is Fable 5.2 better than Opus 5.5?

Per Anthropic's official pages: Fable 5.2 is the most capable model open to all customers, and the announcement says it is a step above Opus 5.5 in almost all the capabilities evaluated in its own benchmark table. The same announcement also says most coding and knowledge tasks should still start on Opus 5.5, and the models overview makes Opus 5.5 the starting point for most workloads. The official guidance is therefore to choose by task, not to replace one model with the other across the board; the benchmarks are vendor-reported.

How far apart are the official prices?

Official prices from Anthropic's pricing page, checked 2026-11-12 and quoted per million tokens: Fable 5.2 is $8 input, $40 output and $0.20 cache reads; Opus 5.5 is $4 input, $20 output and $0.20 cache reads. Cache reads are the same, while Fable 5.2's official input and output prices are higher. For one request of 100,000 input and 10,000 output tokens, the official unit prices work out to $0.40 + $0.20 = $0.60 on Opus 5.5 and $0.80 + $0.40 = $1.20 on Fable 5.2 (a calculation from the official prices, not a vendor-published figure). On QCode, billing is per token with per-model rates on /models.

Which one should I use for coding?

Per the official model-choice guide: start complex agentic coding and enterprise work, multihour autonomous coding agents, large-scale refactoring, complex systems engineering and computer use on Opus 5.5; move to Fable 5.2 when your evals on Opus 5.5 at higher effort still fall short, or when the task is demanding reasoning and long-horizon agentic work. The launch announcement also says Fable 5.2 can act as an Advisor that Haiku, Sonnet or Opus models consult on harder problems.

Can I switch inside the same session?

Yes. Claude Code's official docs say /model with an alias or model name switches immediately during a session, and the choice is written to the model field in your user settings as the default for new sessions. Handle thinking blocks as the official migration guide describes: moving from Fable 5.2 back to Opus 5.5 means Opus cannot read Fable's blocks, so the API drops them without billing and cost or latency can rise on the first turn, while moving from Opus 5.5 to Fable 5.2 keeps the earlier reasoning. The official guide also notes that changing system, tools or earlier messages invalidates Fable 5.2's thinking blocks.

Do both use the same key?

Yes. QCode's /models lists both claude-fable-5-2 and claude-opus-5-5; the same key uses the Anthropic endpoint https://api.qcode.cc/api and you change only the model field. One key covers both Anthropic-compatible and OpenAI-compatible endpoints, and Claude models use the Anthropic one. Billing is per token, with per-model rates on /models.

Are the context and output limits the same?

Yes. The official models overview gives both models 1M tokens of context and 128K tokens of max output, with a reliable knowledge cutoff of June 2026; QCode's /models shows both as 1M / 128K. The one documented output difference is the Batch API: the 300K-output beta list includes Opus 5.5 and does not include Fable 5.2.

Sources

Anthropic official: the Fable 5.2 overview, the Opus 5.5 overview, the models overview, the pricing page, the model-choice guide, the Fable 5.2 migration guide, the Fable 5.2 what's-new page, and the 2026-11-09 launch announcement. Claude Code official: the model-config docs. QCode: qcode.cc/models, qcode.cc/pay-as-you-go and the docs.qcode.cc endpoints page. All fetched and archived on 2026-11-12.

Torn between Fable 5.2 and Opus 5.5?

Create a QCode key and call both models with it, paying per token; Pay-as-you-go Packs give you a one-time quota valid for 7 days.

Create account

Models and live rates View plans