Disclosure: No sponsorship or affiliate links in this article. Benchmarks cited are lab-reported (Anthropic / Cursor). Prices are list prices in USD as of 13 Nov 2026.
Confession first: I haven't used Claude Fable 5.2 yet. I'm not against it. I just haven't worked on any project that Opus 5.5 couldn't handle.
Most of what I build is landing pages, small SaaS features, Apify Actors and bug fixes. That work doesn't need the biggest brain in the room. So this is a practical guide: what Fable 5.2 is, what it costs, what Claude Pro actually shows this week, and where it fits next to the models I already use.
TL;DR
- Do you need Fable 5.2 for vibe coding? Probably not. I haven't needed it yet.
- Price: $8 / $40 per million tokens — 2× Opus 5.5, 4× Sonnet 5.5, and a permanent 20% cut from Fable 5.1 (not 50%).
- My split: Sonnet 5.5 builds, tests and deploys. Opus 5.5 plans and handles auth and payments. Fable only when Opus gets stuck.
- Claude Pro: Fable normally runs on usage credits. Anthropic says it's included up to 50% of your weekly limit until 16 Nov. That's a usage share, not a half-price sale.
What's new in Fable 5.2
From Anthropic's launch post and model docs:
- Released 9 Nov 2026. Model ID
claude-fable-5-2. - $8 / $40 (was $10 / $50 on 5.1). Optional fast mode $16 / $80.
- Up to 30% faster output than Fable 5.1.
- 1M context / 128K max output (same as Opus 5.5 and Sonnet 5.5).
- Available in the Claude apps, Claude Code, Claude Platform, AWS, Google Cloud, Microsoft Foundry, and Cursor's model list.
- Anthropic reports Fable 5.2 ahead of Opus 5.5 on most of its own benches. On FrontierCode 2.0: 49.1% (Xhigh) vs Opus 45.5% (Max). At High effort Fable scored 45.3% — about level with Opus at Max — for roughly half the cost per task (Anthropic's numbers).
OpenRouter and Vercel AI Gateway list the same price. Vercel keeps Fable prompts/completions 30 days for Anthropic safety monitoring. Fable 5.2 is not on OpenRouter's EU in-region routing list right now.
Anthropic bench excerpt
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | vs Opus |
|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% | 32.4% | 45.5% | +3.6 |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | +7.6 |
| OSWorld 2.1 | 89.3% | 83.5% | 87.2% | +2.1 |
| Humanity's Last Exam | 74.2% | 65.6% | 67.7% | +6.5 |
The jump from 5.1 is large. The lead over Opus is smaller: a few points for twice the token price.
Next to other models (price + CursorBench)
| Model | Price in / out | CursorBench 4.0 best |
|---|---|---|
| Claude Fable 5.2 | $8 / $40 | 60.0% · $32.79 (Max) |
| Claude Opus 5.5 | $4 / $20 | 57.8% · $13.43 (Max) |
| Claude Sonnet 5.5 | $2 / $10 | 55.5% · $7.05 (Max) |
| Grok 4.7 (Cursor) | $2 / $6 | 46.3% · $6.01 (Xhigh) |
| Gemini 4 Carbon | $2 / $10 (intro to 30 Nov) | 60.4% · $9.22 (High) |
Takeaway: Sonnet stays within about 5 points of Fable on CursorBench at under a quarter of the cost per task. The best model and the best model for you aren't the same thing.
Do you really need it?
Anthropic itself says start most coding on Opus 5.5 and move up to Fable when Opus at high still falls short (or for long unattended runs). See their model guide.
My default is Sonnet 5.5 on a $20 plan — more building before you hit the limit:
| Model | Job | Price |
|---|---|---|
| Sonnet 5.5 | Build, test, deploy | $2 / $10 |
| Opus 5.5 | Plan, auth, payments | $4 / $20 |
| Fable 5.2 | When Opus is stuck | $8 / $40 |
| Grok 4.7 (Cursor) | Everyday Cursor agent | $2 / $6 |
Sonnet is half Opus's token price and close on most benches. Opus pulls away on hard multi-file work (SWE-bench Pro: Opus 89.9% vs Sonnet 81.3% on Anthropic's table). That's why I use Opus for thinking and Sonnet for doing.
How I work: Opus plans, Sonnet ships
- Plan with Opus 5.5 at High — written plan, tests, definition of done.
- Hand the plan over as a document (switching models drops hidden reasoning).
- Build, test, deploy with Sonnet 5.5.
- Auth and payments go back to Opus.
- Escalate to Fable only when Opus at High is stuck.
Claude Code shortcuts (v2.1.321+): /model sonnet|opus|fable, /effort, /model opusplan. Run claude update first for the Fable 5.2 alias.
Sonnet 5.5 or Grok 4.7 in Cursor?
For everyday features they're close enough that I pick whichever my plan's usage favours:
- Same $2 input; Grok output is $6 vs Sonnet $10.
- CursorBench at High: Sonnet 47.8%, Grok 43.9%. At Medium, Grok leads 41.6% vs 39.2%.
- Not close on Terminal-Bench 4.0 (Anthropic 70.6% for Sonnet vs Cursor 37.6% for Grok) or context (1M vs 256K).
Neither one is a Fable problem.
Fable on Claude Pro vs Max (as of 13 Nov)
Checked against my Pro account and Anthropic's help center (launch offers, Fable on your plan).
| Plan | Fable normally | Launch offer |
|---|---|---|
| Pro / Team standard | Usage credits | Included until 16 Nov up to 50% of weekly limit + one Fable reset |
| Max / Ultra / Team premium | Up to 50% of weekly limits | One to three Fable resets (into 2027) |
| Free | Not available | Not included |
Fine print that matters:
- Pro offer ends 16 Nov, 11:59 pm PT (Tue 17 Nov, 08:59 Budapest). Unused allowance doesn't carry over.
- A Fable reset gives back weekly usage spent on Fable and refills your whole 5-hour window (including Sonnet/Opus). Claim in Settings → Usage; the button isn't in Claude Code yet.
- If you're on Pro and not using Fable, treat Reset for free as a 5-hour refill — claim before it expires.
Is Fable half price this week? No. Permanent API cut is 20%. The “50%” is the share of weekly limit Pro can spend on Fable until 16 Nov.
Cost reality check
Chunky session maths: 400K fresh input + 100K output, no cache:
| Model | Session cost |
|---|---|
| Grok 4.7 (beyond Cursor pool) | $1.40 |
| Sonnet 5.5 | $1.80 |
| Opus 5.5 | $3.60 |
| Fable 5.2 | $7.20 |
| Fable 5.2 fast | $14.40 |
Per token, Fable is 4× Sonnet. Per finished hard task, it can be cheaper than maxing Opus — pay for it when the task is hard, not by default.
FAQ
Worth it for vibe coding? Usually no — Sonnet + Opus cover the loop. Use Fable for long multi-file migrations, big refactors, or agent runs where Opus at High still falls short.
Available on Claude Pro? Yes, normally via usage credits. Launch offer includes Fable up to 50% of weekly limit until 16 Nov. Max/Ultra include Fable up to 50% as standard.
Better than Opus on benches? Anthropic says yes on most of its table; still start everyday coding on Opus.
How do I use it in Claude Code? claude update to v2.1.321+, then /model fable or claude --model fable. /effort defaults to Medium.
Sources: Anthropic Fable / Mythos 5.2. Related notes on Vibe Coder's Life.
