Claude Fable 5.2 vs 5.1: Pricing, Benchmarks, What Changed
Anthropic released Claude Fable 5.2 on November 9, 2026, about two months after Fable 5.1. The upgrade this time is the price and the speed: $8 per million input tokens and $40 per million output tokens (down from $10 and $50), with output that Anthropic says arrives up to 30% faster.
It also shipped with a twin. Claude Mythos 5.2 is the same underlying model with fewer cybersecurity and biology safeguards, and you can only get it through Anthropic's verification programs.
I went through Anthropic's launch post, the Fable 5.2 and Mythos 5.2 model pages, and the fast mode docs, so every number here traces back to a source. If you're weighing Fable 5.2 against Opus 5.5, or deciding whether to move your Fable 5.1 workloads today, this is what just changed in Claude, what it costs, and where it's worth using.
Claude Fable 5.2 at a glance
- Released: November 9, 2026, as the successor to Claude Fable 5.1.
- What it is: Anthropic's model for demanding reasoning and long-horizon agentic work. Mythos 5.2 is the same model for vetted cyber and life-sciences organizations.
- Price: $8 input and $40 output per million tokens, 20% less than Fable 5.1. Fast mode costs $16 and $80.
- Specs: 1M-token context window, 128K max output, text and image input, June 2026 knowledge cutoff.
- Where to use it: Claude's paid apps, Claude Code, the Claude Platform (claude-fable-5-2), Amazon Bedrock, Google Cloud, and Microsoft Foundry.
- The catch: Anthropic still tells most people to start with Opus 5.5 and move up to Fable only when the work calls for it.
What is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic's top generally available model, built for the hardest, most open-ended technical work: agent sessions that run for hours, multistep research, and analysis carried all the way to a finished document, spreadsheet, or deck.
Developers call it with the model ID claude-fable-5-2. It has a 1M-token context window, writes up to 128K tokens per response, takes text and images, and returns text. Adaptive thinking is always on, and you steer how hard it thinks with the effort setting.
Anthropic's launch post makes an unusually modest recommendation about its own flagship. For most coding and knowledge tasks, it tells you to start with Claude Opus 5.5, its "strongest daily driver," then move up to Fable or Mythos, or down to Sonnet 5.5 or Haiku 5.5, as the work requires.
Anthropic also says Fable 5.2 can play two supporting roles in the lineup:
- Advisor: a Haiku, Sonnet, or Opus model consults Fable to help solve harder problems.
- Orchestrator: Fable plans the work and hands simpler tasks to smaller models.
That second role is all over the launch post. Anthropic describes one engineer's experiment where Fable 5.2 coordinated more than 1,000 subagents for eight days to build a prototype operating system that boots on a laptop, starting from an empty folder. (Yes, really.)
Fable 5.2 vs Mythos 5.2: same model, different access
Claude Fable 5.2 and Claude Mythos 5.2 are the same underlying model with the same specs and the same price. The difference is safeguards and who gets to use them.
| 🧠 Claude Fable 5.2 | 🔒 Claude Mythos 5.2 |
|---|---|
| $8 / $40 per 1M tokens | $8 / $40 per 1M tokens |
| Strong cyber and biology safeguards; blocks most security work | Reduced safeguards, for vetted use |
| Open to paid Claude plans, API, cloud platforms | Cyber or Life Sciences Verification Program members only |
| Model ID claude-fable-5-2 | Model ID claude-mythos-5-2 |
| Fast mode available | Fast mode for verified members only |
If Fable 5.2's cybersecurity or biology safeguards block a request in most Claude apps, Anthropic says Claude answers with a fallback model instead: Opus 4.8 for cybersecurity and Opus 5 for biology. On the API, a blocked request comes back as a refusal unless you turn on fallback, which is in beta. Anthropic also says Fable 5.2 produces about 55% fewer safety blocks than Fable 5.1 in early testing across claude.ai, Claude Code, and Cowork. The drop is roughly 60% in cybersecurity and 40% in biology.
Enterprise customers who aren't in a verification program can still reach Mythos 5.2's security skills indirectly, through Claude Security, Anthropic's codebase-scanning product, which now runs on Mythos 5.2.
What's new in Claude Fable 5.2 vs Fable 5.1
Fable 5.2 vs Fable 5.1 comes down to a cheaper bill, a faster response, and fewer false refusals. Here's how they line up on the points that touch your work:
| 🔍 Spec | ⏮️ Fable 5.1 | ⏭️ Fable 5.2 |
|---|---|---|
| Released | September 2026 | November 9, 2026 |
| Input / output per 1M | $10 / $50 | $8 / $40 |
| Cache reads per 1M | $0.25 | $0.20 |
| Cache writes, 5 min / 1 hour | $12.50 / $20 | $10 / $16 |
| Fast mode | Not offered | $16 / $80, up to 2.5x output |
| Output speed | Baseline | Up to 30% faster |
| Context / max output | 1M / 128K | 1M / 128K |
| FrontierCode 2.0 (Xhigh) | 32.4% | 49.1% |
| Remote Labor Index | 17.9% | 41.3% (High) |
A few things stand out once you sit with that table.
It's 20% cheaper on every price line. Input, output, cache reads, and cache writes all drop by the same fifth, and batch requests still get 50% off.
It's faster, with a faster-still option. Output runs up to 30% quicker at standard speed, and the new fast mode raises output speed by up to 2.5x at double the price. One detail from the docs: fast mode speeds up tokens per second, not time to the first token, so it helps long answers more than short ones.
It needs fewer tokens on Medium. Anthropic says that on Medium effort (the default in the Claude apps and Claude Code) Fable 5.2 finishes challenging tasks with fewer tokens than Fable 5.1. Fewer tokens at a lower price is a double discount.
It's better at admitting what it doesn't know. One tester in the launch post put it well: Fable 5.2 "says it doesn't know things instead of hallucinating." Anthropic's own behavioral audit, run on Mythos 5.2, the same underlying model, found lower rates of hallucinated inputs and false claims of finished work. If you've compared earlier Claude models head to head, that honesty is the upgrade you'll feel first.
Claude Fable 5.2 benchmarks
Claude Fable 5.2 benchmarks come straight from Anthropic's launch table, so read them as the vendor's best case. Fable and Opus scores are with safeguards on unless noted, and every model runs at its highest available effort unless noted.
| 📊 Benchmark | 🧠 Fable 5.2 | ⏮️ Fable 5.1 | ⚡ Opus 5.5 | 🟢 GPT-6 Astra | 🔵 Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% (Xhigh) | 32.4% (Xhigh) | 45.5% | 42.8% | 41.1% (Medium) |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| Terminal-Bench-Science 0.1 | 74.6% | 52.6% | 58.7% | 68.1% | 64.8% |
| OSWorld 2.1 (offline set, partial) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| RoboDojo Sim | 64.9% | 27.6% | 39.4% | 29.8% | n/a |
| Humanity's Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | n/a |
| Chartography (no tools) | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
| Remote Labor Index | 41.3% (High) | 17.9% | 21.3% | 20.8% | n/a |
Fable 5.2 takes the top score on most of those rows. Two exceptions are worth flagging, one of them from a benchmark the table leaves out, because an honest table beats a clean sweep:
- Gemini 4 Carbon wins AutomationBench (53.9% vs 51.4%), the test of running business workflows across apps.
- AA-Briefcase is a split: Fable 5.2 posts the higher rubric score (66.1% vs 61.2% for Opus 5.5), but Opus edges it on Elo (1800 vs 1785).
Mythos 5.2, the version with fewer safeguards, carries the security and biology numbers: 76.3% on CyScenarioBench, 89.6% on CVE Rediscovery, and 85.1% on 24-hour protein binder design. Anthropic says Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 on CVE Rediscovery.
On independent data: Artificial Analysis tested Fable 5.2 at Max effort on launch day and scored it 65 on its Intelligence Index, first of the 74 models it tracks, at $26.54 per index task, the most expensive of the 73 it prices.
{{templates}}
The more useful numbers: score per dollar
Anthropic's effort charts show what a given score costs on Fable 5.2:
- Coding: at High effort, Fable 5.2 scores about 45.3% on FrontierCode 2.0, roughly level with Opus 5.5's best (45.5%), for about half the cost per task.
- Agentic search: at Medium effort, it matches Opus 5.5's top WANDR (relaxed score) result for about 40% of the cost per task.
- Expert reasoning: at Low effort, it matches Opus 5.5's top Humanity's Last Exam (with tools) score for less than a tenth of the cost per task.
- Hardware design: at Max effort with tools, it scores 83.5 on BenchCAD 2.1, against 65.3 for Opus 5.5 and 44.4 for Fable 5.1.
That's a real reversal from two months ago, when Artificial Analysis had Opus 5.5 matching Fable 5.1 at default settings for about a third of the cost per task. On hard jobs where you'd otherwise push Opus to Max, Anthropic's charts say Fable 5.2 at a lower effort can now be the cheaper route to the same result. It's the same cost-per-task lens that made Gemini 4 Argon our pick over GPT-6 Astra.
One honest caution: these are Anthropic's own charts, run on Anthropic's own setup. Treat them as a reason to run your own workload before you commit.
Claude Fable 5.2 pricing
Claude Fable 5.2 pricing is per million tokens on the Claude Platform, and the model page lists these rates:
| 💵 Price per 1M tokens | 🧠 Standard | ⚡ Fast mode |
|---|---|---|
| Input | $8.00 | $16.00 |
| Output | $40.00 | $80.00 |
| Cache read | $0.20 | Caching multipliers apply |
| Cache write, 5 min | $10.00 | Caching multipliers apply |
| Cache write, 1 hour | $16.00 | Caching multipliers apply |
| Batch API | 50% off input and output | Not available |
Fast mode runs at the same multiplier across the full context window.
What one agent session costs
Here's the same illustrative session we priced for Opus 5.5 and Fable 5.1 in September: 8M cached tokens read back, 400K fresh tokens written to a five-minute cache, and 150K tokens of output, at list rates.
| 🧾 Line item | ⚡ Opus 5.5 | ⏮️ Fable 5.1 | ⏭️ Fable 5.2 |
|---|---|---|---|
| 8M cache reads | $1.60 | $2.00 | $1.60 |
| 400K cache writes (5 min) | $2.00 | $5.00 | $4.00 |
| 150K output tokens | $3.00 | $7.50 | $6.00 |
| Session total | $6.60 | $14.50 | $11.60 |
These are toy numbers, so don't quote them to finance. Fable 5.2 lands about 20% below Fable 5.1 and roughly 1.8x Opus 5.5 on the same tokens. Cache reads now cost the same on Fable 5.2 and Opus 5.5 ($0.20). They're the smallest line in this toy session, but Anthropic's Opus 5.5 launch post says cache reads make up the majority of agentic and coding work costs, so on a real agent bill that line no longer favors Opus.
On a Claude subscription
For Fable 5.2's launch, Anthropic is giving eligible paid subscribers extra Fable usage:
- Max and Ultra plans and premium Team seats get one to three Fable usage resets, depending on how long you've subscribed, if paid billing was in place before November 9. The resets last until March 8, 2027.
- Pro plans and standard Team seats get Fable included for up to 50% of weekly usage limits through November 16, 2026, plus one Fable reset.
Read that as a launch promotion. Outside it, Fable 5.2 works like Fable 5.1: Max and Ultra plans, premium Team seats, and premium seats on legacy seat-based Enterprise plans include Fable for up to 50% of weekly limits, while Pro and standard seats pay for it with usage credits. If you're comparing tiers in the first place, our guide to Claude's pricing walks through what each plan actually gets you.
How to access Claude Fable 5.2
Claude Fable 5.2 is available on every platform Anthropic lists, as of launch day. The API model ID is claude-fable-5-2.
| 🧭 Where | 🛠️ How to get it | ⚠️ Notes |
|---|---|---|
| Claude apps | Model picker on eligible paid plans | Launch offer on Pro and standard Team seats through Nov 16 |
| Claude Code | /model or claude --model claude-fable-5-2 | Fast mode supported; default effort Medium |
| Claude Platform (API) | claude-fable-5-2 | Default effort High; 30-day data retention |
| Amazon Bedrock | global.anthropic.claude-fable-5-2 | Also on the Claude Platform on AWS |
| Google Cloud | claude-fable-5-2 | Anthropic or Google managed; fast mode on both |
| Microsoft Foundry | claude-fable-5-2 | Anthropic-hosted deployments |
If you work in Claude Code, update to version 2.1.321 or later before you switch. You can flip models mid-session with /model and keep your conversation when you do. Fable is a natural fit for AI coding agents that run for hours, where Fable 5.2 at a lower effort can match Opus 5.5 at Max for less per task.
Five things to know before you switch API traffic
- Data retention: Fable 5.2 uses 30-day data retention for safety monitoring unless your organization is expressly authorized for zero data retention.
- Thinking blocks: if a router, retry, or fallback moves a conversation from Fable 5.2 to another model, Fable 5.1 and Opus 5.5 included, the API drops Fable 5.2's thinking blocks, so the turns after the switch run without that reasoning.
- Computer use: the older computer_20251124 tool returns a 400 error on every platform except Amazon Bedrock, so move to the computer use toolset first.
- Preserved thinking: editing earlier turns now invalidates thinking blocks for every account, part of Anthropic's anti-distillation work. Check any integration that rewrites conversation history.
- Advisor tool: a Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor, so update the advisor model in the same change.
Sampling parameters and forced tool choice already work this way on Fable 5.1 and Opus 5.5, so that request code carries over. Anthropic commits to keeping Fable 5.2 available until at least November 9, 2027.
Where Claude Fable 5.2 falls short
Fable 5.2 is a genuine step up, and it still has rough edges worth knowing before you move production traffic:
- It's twice Opus 5.5's token price. For everyday chat, writing, and routine coding, Anthropic itself points you to Opus 5.5 first.
- It's not a low-latency model. The model page labels its latency "Moderate," the same as Opus 5.5, while Sonnet 5.5 is "Fast." Fast mode speeds up output, not the wait for the first token.
- Security work gets blocked. Penetration testing, exploit generation, and most research biology are blocked on Fable 5.2. Most Claude apps hand those requests to a fallback model, while the API returns a refusal unless you set fallback up. That work needs Mythos 5.2 and a verification program.
- No zero data retention by default. Teams with strict retention rules may need to stay on Opus 5.5 until they can get Enterprise Frontier Safeguards, now in private beta. It runs the safety monitoring on prompts and responses stored in their own infrastructure, so it works with zero data retention.
- Most of the benchmarks are Anthropic's own. The one independent read so far, Artificial Analysis, ranks it first on intelligence and last on cost per task, and Gemini 4 Carbon still leads the one business-workflow test in the table.
- It often suspects it's being tested. Anthropic reports that Fable 5.2 voices suspicion that it's being evaluated more often than any of Anthropic's recent models, which makes safety evals harder to read. Anthropic still rates the high-stakes risk as low.
Is Claude Fable 5.2 worth it? A quick decision guide
Whether Fable 5.2 earns its premium comes down to how hard your work actually is.
Choose Fable 5.2 if you:
- Run long, open-ended agent work, big refactors, or migrations where a failed run costs more than the tokens.
- Already use Fable 5.1, since it's cheaper and faster with little reason to stay behind.
- Orchestrate teams of smaller models and want a stronger planner sitting on top.
- Do research where a model saying "I don't know" beats a confident guess.
- Want to try it before November 16, while the launch offer adds extra usage on Pro and standard Team seats.
Stick with Opus 5.5 (or go lower) if you:
- Mostly chat, write, or do routine coding, where Anthropic recommends Opus 5.5 first.
- Need zero data retention or the lowest latency.
- Watch every dollar on high-volume API traffic, where a cheaper model may be the better buy.
The simplest test is the one Anthropic's own advice points to: keep Opus 5.5 as your default, keep a short list of the tasks it fails at higher effort, and rerun only those on Fable 5.2 while the extra usage lasts. For the full head-to-head on that call, see how Fable 5.2 and Opus 5.5 stack up.
Choosing your model for each task
Once you've settled on a model, the work itself still has to get done across your inbox, calendar, and CRM, which is the part no benchmark table covers.
Lindy is an AI assistant you text, and it's model agnostic, so you pick whichever major model fits each task. With Lindy, teams can:
- Text an assistant that lives in Slack and hand it work in plain English.
- Choose the model per task instead of locking your whole workflow to one.
- Hand off real work: inbox triage, meeting scheduling, follow-ups, and CRM updates.
- Keep a human in the loop, since Lindy can ask for your OK before it sends or schedules anything.
Before you plan anything around Fable 5.2, it's worth checking which models your workspace lists. Try Lindy free and point it at the work you'd rather not do by hand.
{{cta}}
Frequently asked questions
When was Claude Fable 5.2 released?
Claude Fable 5.2 was released on November 9, 2026, alongside Claude Mythos 5.2. It follows Claude Fable 5.1 and Mythos 5.1, which launched in September 2026.
How much does Claude Fable 5.2 cost?
Claude Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1. Cache reads cost $0.20 per million, batch requests are half price, and fast mode costs $16 and $80.
What is the difference between Claude Fable 5.2 and Claude Mythos 5.2?
The main difference between Claude Fable 5.2 and Claude Mythos 5.2 is safeguards and access. Both are the same model at the same price, but Fable has strong cybersecurity and biology safeguards and is available on all paid Claude plans and the API, while Mythos has fewer safeguards and is limited to organizations in Anthropic's Cyber and Life Sciences Verification Programs.
Is Claude Fable 5.2 better than Opus 5.5?
Yes, Claude Fable 5.2 beats Opus 5.5 on most of Anthropic's published benchmarks, including FrontierSWE v2 (69.9% vs 62.3%) and Humanity's Last Exam with tools (74.2% vs 67.7%). It also costs twice as much per token, and Anthropic still recommends Opus 5.5 as the starting point for most work.
Is Claude Fable 5.2 free?
No, Claude Fable 5.2 isn't free. It's available on eligible paid Claude plans and billed per token on the Claude Platform and cloud platforms, though Pro plans and standard Team seats get Fable included through November 16, 2026.
What is Claude Fable 5.2's context window?
Claude Fable 5.2 has a 1M-token context window and can write up to 128K tokens per response. Its reliable knowledge cutoff is June 2026.
Does Claude Fable 5.2 have a fast mode?
Yes, Claude Fable 5.2 has an optional fast mode that raises output speed by up to 2.5 times for $16 per million input tokens and $80 per million output tokens. It's available on Claude Code, the Claude Platform, Google Cloud, and Anthropic-hosted deployments on Microsoft Foundry, but not with the Batch API.
Is Fable 5.1 being retired?
No, Fable 5.1 isn't being retired yet. Anthropic lists it as active, with retirement no sooner than September 1, 2027.
