Fable 5.2 vs Opus 5.5 in Claude Code: $2 vs $1 a task, 6 vs 1–2 points of the 5-hour limit
On our Claude Code task, Fable 5.2 cost about twice as much as Opus 5.5 at API prices, and on our Max plan it moved the 5-hour limit about 4.5 times as far. Keep Opus 5.5 as your default and switch to Fable 5.2 for the tasks that need it.
The 30-second version
On the same Claude Code task, each Fable 5.2 run moved our Max 5x plan’s 5-hour limit by 6 points; each Opus 5.5 run moved it by 1 or 2.
- Price: about 2x. A Fable 5.2 run cost about $2 at list prices, an Opus 5.5 run about $1. Both wrote similar amounts of output.
- 5-hour limit: about 4.5x. Three Fable 5.2 runs took 18 points, three Opus 5.5 runs 4.
- Both finished. All six runs passed the test suite. Fable 5.2 took 3.4 minutes on average, Opus 5.5 2.9.
- What to do. Keep Opus 5.5 as the default, switch to Fable 5.2 with
/modelfor the work that needs it, and check/usagethe first few times.
Same task, three runs each
Each run started from a fresh copy of the same public repository and asked Claude Code 2.1.322 to add a new data source, with offline tests and documentation, then run the test suite. Both models ran at their default effort, medium, and we alternated the order:
| Run | Model | Cost at list prices | Time | Turns | 5-hour limit |
|---|---|---|---|---|---|
| 1 | Opus 5.5 | $1.01 | 3.1 min | 19 | 23% → 24% |
| 2 | Fable 5.2 | $1.82 | 2.8 min | 11 | 24% → 30% |
| 3 | Fable 5.2 | $2.11 | 3.8 min | 16 | 30% → 36% |
| 4 | Opus 5.5 | $0.98 | 2.6 min | 17 | 36% → 38% |
| 5 | Opus 5.5 | $1.02 | 3.0 min | 20 | 38% → 39% |
| 6 | Fable 5.2 | $2.00 | 3.6 min | 13 | 39% → 45% |
Output counts were close. Each run wrote 15,000 to 19,000 output tokens and 59,000 to 75,000 tokens to the cache. Opus 5.5 took more turns and read more from the cache, but cache reads cost $0.20 per million tokens on both models. Every other Fable 5.2 price is twice Opus 5.5’s: $8 against $4 per million input tokens, $40 against $20 per million output tokens.
Anthropic says Fable uses limits faster, not how much
On our task, Fable 5.2 moved the 5-hour limit about 4.5 times as far as Opus 5.5. Anthropic’s help center says only that Fable models use your limits “faster than other Claude models”, and Claude Code 2.1.322’s startup notice that Fable 5.2 “draws down usage faster than Opus 5.5”.
The weekly limits moved far less. Across all our runs, the overall weekly meter went from 10% to 11%, and the separate weekly Fable meter from 17% to 19%.
When Fable 5.2 is worth it
Anthropic recommends starting with Opus 5.5 “for most coding and knowledge tasks” and moving up to Fable for “the most ambitious, open-ended technical work”. On FrontierCode 2.0, its announcement says, Fable 5.2 at high effort matched Opus 5.5’s best score “for roughly half the cost per task”. At default effort on our task, Fable 5.2 cost twice as much, and on our plan the gap was wider.
How we measured
- Runs. Six headless runs (
claude -p) of Claude Code 2.1.322 on 12 November 2026, each in a fresh clone of the public lastdays-skill repository at the same commit, with the same prompt and permissions skipped. - Cost and tokens. Claude Code’s own per-run totals at list prices (
total_cost_usdandmodelUsagein its JSON output). - Limits. One Max 5x account. We read the plan’s usage figures, the ones
/usageshows, between runs; they come in whole points. Other sessions on the account were idle, apart from a few small requests from this write-up’s own session.
What we didn’t test
- Other tasks and efforts. One medium-sized task at default effort. Harder tasks, or Fable 5.2 at high effort, may favor it more.
- Code quality beyond the test suite passing.
- Other plans: Pro, Max 20x and Team.
- Precision. A move of 6 points on a whole-point meter can mean anything from just over 5 to just under 7.
Companion reading
- Claude Code usage limits — how the 5-hour and weekly limits work.
- Claude Opus 5 vs Fable 5 — the earlier comparison, before the 5.2 and 5.5 releases.
- Choosing a model and effort in Claude Code — what effort levels change.
- Telemetry off in Claude Code? Its Fable row still picks Fable 5.1 — if your
/modelpicker doesn’t list Fable 5.2.
Sources
- Anthropic — Claude Fable 5.2 and Mythos 5.2, 9 November 2026.
- Anthropic — Pricing. Retrieved 12 November 2026.
- Claude Help Center — Claude Fable models on your plan. Retrieved 12 November 2026.
- Anthropic — Claude Code changelog, 2.1.321 entry.
FAQ
Does Fable 5.2 use more of my Claude Code limit than Opus 5.5? In our test, yes. On a Max 5x plan, three runs of the same Claude Code task moved the 5-hour limit by 18 points on Fable 5.2 and 4 on Opus 5.5, although Fable 5.2 cost only about twice as much at API prices.
How much does a Claude Code task cost on Fable 5.2? Our task cost $1.82 to $2.11 per run on Fable 5.2 at list prices, against $0.98 to $1.02 on Opus 5.5. Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, twice Opus 5.5’s $4 and $20; cache reads cost $0.20 per million on both.
Is Fable 5.2 cheaper than Fable 5.1? At API prices, yes: $8 and $40 per million input and output tokens, against $10 and $50 for Fable 5.1.
Should I make Fable 5.2 my default model in Claude Code? Anthropic recommends starting with Opus 5.5 for most coding tasks, and in our test Fable 5.2 moved the 5-hour limit about 4.5 times as far per task. Switch to it with /model for the tasks that need it.
Was this helpful?