By the toolist.ai editors. Researched from public sources, not hands-on tested by toolist.ai. Updated November 19, 2026; prices checked November 19, 2026.

Claude Fable 5.2 vs Claude Opus 5.5

Verdict

Start with Claude Opus 5.5, as Anthropic advises, and move up to Claude Fable 5.2 when a task needs more. Fable 5.2 costs twice as much ($8 and $40 per million tokens against $4 and $20) and leads on the Artificial Analysis Intelligence Index (65 against 59 at max effort, 62 against 48 at each model's default effort), HLE-Diamond (80% against 68%) and Terminal-Bench-Science (73% against 59%). The two tie at 60% on Terminal-Bench 4.0, and cache reads cost $0.20 per million tokens on both.

Verdict based on published prices, limits and benchmarks, not on our own tests.

Choose Claude Fable 5.2 if

  • Demanding reasoning and long-horizon agentic work, which Anthropic names Fable 5.2 for
  • Hard research questions: 80% against 68% on HLE-Diamond
  • You want the higher score at default settings: 62 against 48 on the Intelligence Index

Choose Claude Opus 5.5 if

  • Most workloads, where Anthropic recommends starting, at half the token price: $4 and $20 per million tokens
  • Terminal coding, where the two tie at 60% on Terminal-Bench 4.0
  • You need zero data retention: Opus 5.5 is available under it, while Fable 5.2 requires 30-day retention
  • Fast mode at a lower price: $8 and $40 per million tokens against $16 and $80

Price

Input costs 2 times as much on Claude Fable 5.2 as on Claude Opus 5.5. Output costs 2 times as much on Claude Fable 5.2 as on Claude Opus 5.5.

Per million tokensClaude Fable 5.2Claude Opus 5.5
Input$8 source$4 source
Cached input$0.20 source$0.20 source
Output$40 source$20 source

Claude Fable 5.2: Cache writes cost $10 per million tokens (5-minute cache) or $16 (1-hour cache). The Batch API costs half. There is no long-context surcharge. Fast mode, with up to 2.5 times faster output, costs $16 in and $80 out and is not available with the Batch API.

Claude Opus 5.5: Cache writes cost $5 per million tokens (5-minute cache) or $8 (1-hour cache). The Batch API costs half. There is no long-context surcharge. Fast mode costs $8 in and $40 out.

Worked example: what a month costs

We priced three monthly workloads at the standard rates. The provider does not publish these figures.

WorkloadClaude Fable 5.2Claude Opus 5.5
Support chatbot
100,000 replies, each 3,000 input and 400 output tokens, no caching (300M in, 40M out)
$4,000$2,000
Document Q&A
5,000 questions over the same 60,000-token document set (90% cache hits), 800 output tokens each (300M in, 4M out)
$454$254
Coding agent
200 tasks, each 1.5 million input tokens (90% cache hits) and 50,000 output tokens (300M in, 10M out)
$694$374

Price your own workload

ModelCost a month
Claude Fable 5.2 $600
Claude Opus 5.5 $300

Standard API prices, no batch or priority discounts. Cached input is billed at the cache-read price; cache-write charges are not included. Reasoning models bill their thinking as output tokens.

Limits and release dates

Claude Fable 5.2Claude Opus 5.5
ReleasedNovember 9, 2026 sourceSeptember 22, 2026 source
Context window1,000,000 tokens source1,000,000 tokens source
Maximum output128,000 tokens source128,000 tokens source
Knowledge cutoffJune 2026 sourceJune 2026 source
Input and outputText and images in; text out sourceText and images in; text out source
API model nameclaude-fable-5-2claude-opus-5-5

Benchmarks

Scores as published by the provider or measured by Artificial Analysis; higher is better. A dash means no score was published.

BenchmarkClaude Fable 5.2Claude Opus 5.5
Artificial Analysis Intelligence Index
max effort
65.0 source59.0 source
Intelligence Index at everyday effort 62.0 (high effort, the API default) source48.0 (medium effort, the API default) source
Terminal-Bench 4.0
max effort
60.0% source60.0% source
HLE-Diamond
max effort
80.0% source68.0% source
Terminal-Bench-Science 0.1
max effort
73.0% source59.0% source
Output speed 107.4 tokens/s (high effort, Artificial Analysis) source104.9 tokens/s (medium effort, Artificial Analysis) source

Benchmark and speed data from Artificial Analysis where linked.

Read more

Sources