SaaS · Analysis
Claude Fable 5.2 Pricing: 20% Cheaper, Still 2x Opus
Anthropic cut Claude Fable 5.2 to $8 input and $40 output per million tokens, 20% below Fable 5.1 but still twice Opus 5.5. Here is the cost math and when moving up to Fable pays off.
Metarix Editorial · Nov 10, 2026 · 9 min read

On this page
Anthropic released Claude Fable 5.2 on November 9, 2026, and cut its API price by 20%: $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. If you already run Fable, switching the model ID is the easiest saving you will get this quarter. If you run Opus 5.5, Fable is still twice the per-token price, so the real question is whether it finishes your hardest tasks with fewer tokens or fewer retries.
Here is what changed, what it costs on real workloads, and how to decide between Fable 5.2, Opus 5.5 and the cheaper tiers.
What Anthropic shipped
Anthropic's launch post introduces two models, Claude Fable 5.2 and Claude Mythos 5.2, which it says are the same underlying model with different safeguards. Fable is available to everyone; Mythos is limited to organizations vetted through Anthropic's cyber and life sciences verification programs. For most builders, Fable is the one that matters.
- Lower price: $8 input and $40 output per million tokens on the Claude API, per Anthropic's pricing docs.
- Faster output: Anthropic says standard generation is up to 30% faster than Fable 5.1.
- New fast mode: up to 2.5 times faster token generation at $16 input and $80 output per million tokens.
- Where to get it: the Claude Platform as
claude-fable-5-2, Claude apps and Claude Code on paid plans, plus Amazon Web Services, Google Cloud and Microsoft Azure. - Fewer false blocks: Anthropic reports about 55% fewer safety-classifier blocks than Fable 5.1 in early testing across its apps.
The new price list, side by side
The 20% cut applies across the board, including cache writes and cache reads, so the same mix of traffic simply costs a fifth less. Prices as of November 10, 2026, from Anthropic's API pricing documentation.
Scroll the table horizontally on smaller screens.
| Model | Input | Output | Cache hit | Batch input / output |
|---|---|---|---|---|
| Claude Fable 5.2 | $8 | $40 | $0.20 | $4 / $20 |
| Claude Fable 5.2, fast mode | $16 | $80 | Multipliers apply on top | Not available |
| Claude Fable 5.1 | $10 | $50 | $0.25 | $5 / $25 |
| Claude Opus 5.5 | $4 | $20 | $0.20 | $2 / $10 |
| Claude Sonnet 5.5 | $2 | $10 | $0.10 | $1 / $5 |
Two details are easy to miss. First, Fable's cache hits are billed at 2.5% of base input, against 5% on Opus 5.5, so a cache hit now costs the same $0.20 per million tokens on both models. In a long agent loop that rereads a big cached prefix every turn, that narrows the gap between them. Second, the 1 million token context window is billed at standard rates, with no long-context surcharge.
Some modifiers stack on top. US-only inference through inference_geo adds a 1.1x multiplier, and Amazon Bedrock and Google Cloud's Google-managed offering set their own prices, so check those pages if you buy Claude through a cloud. Our Claude API vs Bedrock vs Vertex AI comparison covers those trade-offs.
What it costs on real workloads
Per-token prices only matter once you plug in a workload shape. The table below works out three common patterns at list prices. It is our arithmetic, not vendor data, and it assumes the same token counts on every model, which will not hold in practice because models use different numbers of reasoning and output tokens for the same task.
Scroll the table horizontally on smaller screens.
| Workload | Fable 5.1 | Fable 5.2 | Opus 5.5 | Fable 5.2 Batch |
|---|---|---|---|---|
| Big one-shot job: 200K input, 20K output | $3.00 | $2.40 | $1.20 | $1.20 |
| Agent turn: 150K cached prefix read, 10K new input, 5K output | $0.39 | $0.31 | $0.17 | n/a (interactive) |
| Code review: 40K input, 4K output | $0.60 | $0.48 | $0.24 | $0.24 |
The takeaways are simple. Anyone already on Fable 5.1 saves 20% for the same traffic. Opus 5.5 is still about half the price for the same token counts. And Fable through the Batch API now costs the same as interactive Opus, which makes Fable a realistic choice for overnight work such as large refactor plans, repository audits and evaluation runs that don't need an answer in seconds.
Why per-token price is not the whole story
The more interesting claim in Anthropic's launch is about cost per task, not cost per token. On FrontierCode 2.0, which scores whether a model's code changes would be merged by open-source maintainers, Anthropic says Fable 5.2 at High effort scored 45.3%, about level with Opus 5.5's best score of 45.5%, for roughly half the cost per task. At its highest effort setting, Fable 5.2 reached 49.1%, which Anthropic calls the best of any released model it tested.
Scroll the table horizontally on smaller screens.
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% (Xhigh effort) | 32.4% (Xhigh effort) | 45.5% |
| FrontierSWE v2 (agentic coding) | 69.9% | 56.3% | 62.3% |
| OSWorld 2.1, offline set, partial (computer use) | 89.3% | 83.5% | 87.2% |
| Humanity's Last Exam, with tools | 74.2% | 65.6% | 67.7% |
Independent numbers are starting to arrive. Artificial Analysis' changelog listed Fable 5.2 at 65 on its Intelligence Index at Max effort and 56 at Medium effort on November 9, 2026. Treat both sets of figures with the usual caution: vendor benchmarks use vendor harnesses, and aggregate indexes hide the task mix. Our guide to reading LLM leaderboards explains what to check before trusting a chart.
The practical point is that a pricier model can be cheaper per finished task if it needs fewer tokens, fewer turns or fewer retries. That only shows up if you measure cost per successful task on your own workload, as we described in our cost per successful task guide.
Fable 5.2 or Opus 5.5: how to choose
Anthropic itself recommends starting most coding and knowledge work on Opus 5.5 and moving up to Fable only when the job needs it. That is sensible advice, and the price cut doesn't change it. It does lower the bar for when moving up is worth it.
Scroll the table horizontally on smaller screens.
| Situation | Start with | Why |
|---|---|---|
| Everyday coding, chat, docs, most agent steps | Opus 5.5 | Half the per-token price; Anthropic's recommended default |
| Long, ambiguous projects, big migrations, unfamiliar systems | Fable 5.2 | Vendor results show the biggest gains on hard agentic work |
| Orchestrating a team of agents | Fable 5.2 as planner, cheaper models as workers | Pay top rates only for the planning and review steps |
| Overnight audits, refactor plans, eval runs | Fable 5.2 via Batch API | Same list price as interactive Opus 5.5 |
| Latency-sensitive interactive use | Opus 5.5, or Fable 5.2 fast mode if quality is critical | Fast mode doubles Fable's price |
| Security tooling, exploit or pentest workflows | Not Fable | Fable's safeguards block most of these tasks; Mythos needs verification |
The last row matters for anyone building security products. Anthropic says Fable 5.2's classifiers block high-risk cyber tasks such as penetration testing and exploit generation, and that blocked queries are answered by an older model instead. If your product touches that area, test it before you switch, or apply for Mythos access.
How to switch, and what to check
For teams already on Fable 5.1, the move is a model ID change. Here is an illustrative call with Anthropic's Python SDK; it was not run in this session, and you should keep your existing parameters for effort, tools and caching.
# Illustrative only: swap the model ID, keep the rest of your requestimport anthropicclient = anthropic.Anthropic() # reads ANTHROPIC_API_KEY from the environmentmsg = client.messages.create( model="claude-fable-5-2", # was the Fable 5.1 model ID max_tokens=4096, messages=[{"role": "user", "content": "Plan the migration of our billing service."}],)print(msg.usage) # compare input, output and cache tokens with your 5.1 baseline
- Rerun your evals. A new model can change output length, tool-call patterns and refusal behavior. Compare cost per successful task, not just the bill.
- Watch the cache. Cache hits are now $0.20 per million tokens. If your agent reuses a large prefix, check your hit rate; our prompt caching break-even guide shows how to work out whether caching pays.
- Mind the tokenizer if you are coming from an older model. Anthropic's pricing docs say Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text, so a jump from Sonnet 4.6 or earlier will not be a like-for-like price comparison.
- Check your cloud. On Bedrock or Google Cloud's Google-managed service, prices and fast-mode availability are set by the provider, so confirm before you budget.
Subscribers get a short launch window too. Anthropic's launch offers page says Pro plans and standard Team seats get all Fable models included for one week, up to 50% of weekly limits, plus one Fable limit reset, while Max, Ultra and premium seats get one to three Fable resets that expire on March 8, 2027. Resets only go to accounts with paid billing in place before 10:00 am PT on November 9, 2026. That is a cheap way to try Fable in Claude Code on a real task before you change an API budget.
Verdict
If you pay for Fable 5.1 today, switch to Fable 5.2 now: it is 20% cheaper at the same traffic, and Anthropic reports better results across coding, computer use and reasoning. If you run Opus 5.5, don't switch by default; it remains half the per-token price and Anthropic's own recommended starting point. Instead, route only your hardest, longest tasks to Claude Fable, try the Batch API for anything that can wait, and measure cost per finished task rather than cost per token.
For founders, the bigger signal is direction. Anthropic cut the price of its top model one generation after launch while widening the gap on hard agentic work. Budget for frontier rates on the few steps that need them, and keep your routing flexible, because the cheapest model for a task keeps changing.
Sources
- Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 (Nov 9, 2026)
- Anthropic: Claude API pricing documentation (accessed Nov 10, 2026)
- Anthropic: Claude pricing page (accessed Nov 10, 2026)
- Claude Help Center: Claude Fable 5.2 launch offers (accessed Nov 10, 2026)
- Artificial Analysis: Changelog (entries dated Nov 9, 2026)
- VentureBeat: Anthropic releases Claude Fable 5.2 and Mythos 5.2 (Nov 9, 2026)
Tools covered in this article
Frequently asked questions
- How much does Claude Fable 5.2 cost?
- As of November 10, 2026, Anthropic lists Claude Fable 5.2 at $8 per million input tokens and $40 per million output tokens on the Claude API. Cache hits cost $0.20 per million tokens, the Batch API halves the price to $4 and $20, and the optional fast mode costs $16 and $80.
- Is Claude Fable 5.2 cheaper than Fable 5.1?
- Yes. Fable 5.1 costs $10 per million input tokens and $50 per million output tokens, so Fable 5.2 is 20% cheaper. Cache writes and reads fell by the same proportion, so an identical traffic mix costs about 20% less.
- Should I use Claude Fable 5.2 or Opus 5.5?
- Anthropic recommends starting most coding and knowledge work on Opus 5.5, which costs half as much per token, and moving to Fable 5.2 for the hardest, most open-ended tasks. Anthropic reports that Fable 5.2 at High effort matched Opus 5.5's best FrontierCode 2.0 score for roughly half the cost per task, so test both on your own workload.
- What is the model ID for Claude Fable 5.2?
- The model ID on the Claude Platform is claude-fable-5-2. Fable 5.2 is also available on Amazon Web Services, Google Cloud and Microsoft Azure, where the provider may set its own prices.