Skip to main content
BenchLM
Data

Claude Fable 5.2 vs Claude Haiku 5.5

Updated November 9, 2026. Rank says Claude Fable 5.2 is ahead. Price, access, and your workload can each overturn that. Public scores include evidence status and uncertainty.

Share or export
Share on XLinkedInAPI/MCP

Decision reading

Claude Fable 5.2 has the higher public point estimate, 91.19 versus 66.12. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. 8 results are shared. Category rows resting on Estimated evidence or different benchmark sets are marked directional and do not name a winner.

Model A
Anthropic logo
Claude Fable 5.2

Anthropic

91.19/100

Estimated · Public rank #1

Conditional range 76.8–100.0

Model B
Anthropic logo
Claude Haiku 5.5

Anthropic

66.12/100

Supported · Public rank #27

90% interval 57.6–74.6

Shared results
8
Claude Fable 5.2 only
22
Claude Haiku 5.5 only
13
Like-for-like categories
2 / 8
Estimated: Claude Fable 5.2 · Supported: Claude Haiku 5.5. Conditional ranges do not establish rank confidence.How the comparison works

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Chat turn cost

    1K fresh input + 500 output tokens

    Claude Haiku 5.5

    Claude Haiku 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Claude Haiku 5.5

    Claude Haiku 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates
Show secondary and unsupported calls

  • Repository review cost

    50K fresh input + 3K output tokens

    Claude Haiku 5.5

    Claude Haiku 5.5 has the lower estimated token cost for this stated workload. Costs use the listed standard API rates.

    Confidence: listed-rates

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

    Claude Fable 5.2 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like.

    Confidence: limited

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    Claude Fable 5.2 is scored on Estimated evidence for agentic, so the reading is directional rather than like-for-like.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    No clear pick

    The documented context windows are equal.

    Confidence: documented

Which one for a specific job

Choose a job from the LLM Selector's task catalog to see the category row it rests on, under the same basis rules as the table below. A directional row stays directional; choosing a job never creates a winner.

The same task catalog as the LLM Selector. Each job names the evidence surface it rests on; nothing here adds a new score.

86.9Claude Fable 5.251.7Claude Haiku 5.5

Directional only · BenchAlign v5.8

Claude Fable 5.2 has the higher coding point estimate. Conditional score ranges do not establish rank confidence.

Coding scores combine specific tasks and setups. Match the editor, harness, and effort to your workflow.

Open the coding leaderboardShortlist models for this job
Same basis rules as the category table below

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

3 categories rest on Estimated evidence or different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

Bars run 0–100 on each benchmark’s normalized display scale

Category results, on a stated basis

Each row shows the public-lane category score for both models: the BenchAlign v5.8 lane for agentic, coding, reasoning, multimodal, and knowledge, and the weighted public rows elsewhere, with the same rank each model page reports. A row is like-for-like only when both scores rest on Supported evidence or the same weighted set. Directional and not-comparable rows remain visible, but they never receive a winner.

Reasoning

Like-for-like
Claude Fable 5.2
90.7
Supported · #1/157
Claude Haiku 5.5
58.1
Supported · #38/157
Basis
BenchAlign v5.8 lane · 3 vs 4 public rows
Reading
Claude Fable 5.2 leads

Knowledge

Like-for-like
Claude Fable 5.2
97.6
Supported · #1/179
Claude Haiku 5.5
63.4
Supported · #38/179
Basis
BenchAlign v5.8 lane · 3 vs 1 public rows
Reading
Claude Fable 5.2 leads

Agentic

Directional only
Claude Fable 5.2
90.7
Estimated · #1/132
Claude Haiku 5.5
60.7
Supported · #25/132
Basis
BenchAlign v5.8 lane · 8 vs 7 public rows
Reading
Directional only

Coding

Directional only
Claude Fable 5.2
86.9
Estimated · #1/158
Claude Haiku 5.5
51.7
Supported · #43/158
Basis
BenchAlign v5.8 lane · 8 vs 4 public rows
Reading
Directional only

Multimodal

Directional only
Claude Fable 5.2
89.8
Estimated · #1/67
Claude Haiku 5.5
60.7
Supported · #35/67
Basis
BenchAlign v5.8 lane · 6 vs 3 public rows
Reading
Directional only

Multilingual

Not comparable
Claude Fable 5.2
Not ranked
Claude Haiku 5.5
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Instruction following

Not comparable
Claude Fable 5.2
Not ranked
Claude Haiku 5.5
Not ranked
Basis
Provisional lane · 0 vs 0 weighted rows
Reading
Not comparable

Math

Not comparable
Claude Fable 5.2
Not ranked
Claude Haiku 5.5
75.6
Unranked · 2 rankable rows
Basis
Provisional lane · 0 vs 2 weighted rows
Reading
Not comparable

Ranks count the models scored in each category’s lane, so the agentic, coding, reasoning, multimodal, and knowledge denominators (BenchAlign v5.8) differ from the provisional-lane categories. Unranked scores sit on the provisional lane’s scale but never name a winner. Like-for-like rows within 0.5 points read as a practical tie.

Supported evidence per lane · bars run 0–100Methodology

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Claude Fable 5.2
$0.028
Fits in one request
Claude Haiku 5.5
$0.00035
Fits in one request

Claude Haiku 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Repository review

50K fresh input + 3K output tokens

Claude Fable 5.2
$0.52
Fits in one request
Claude Haiku 5.5
$0.0065
Fits in one request

Claude Haiku 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Claude Fable 5.2
$0.6
Fits in one request
Claude Haiku 5.5
$0.009
Fits in one request

Claude Haiku 5.5 has the lower modeled cost

Costs use the listed standard API rates.

Cached input falls back to the list input rate only where a cached rate is unpublished

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Claude Fable 5.2

$0.2 per 1M cached input tokens

Claude API pricing

Claude Haiku 5.5

$0.01 per 1M cached input tokens

Claude API pricing

Reasoning profile

Claude Fable 5.2

Reasoning

Claude Haiku 5.5

Reasoning

Weight access

Claude Fable 5.2

Proprietary

Claude Haiku 5.5

Proprietary

License

Claude Fable 5.2

Proprietary

Claude Haiku 5.5

Proprietary

Release date

Claude Fable 5.2

2026-11-09

Claude Haiku 5.5

2026-10-07

If you already use one of these models

Deployment change
Both entries list Anthropic as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
Quality signal
Claude Fable 5.2 has the higher public point estimate, 91.19 versus 66.12. Their conditional score ranges do not overlap. These ranges do not establish rank confidence.
Workload cost
Repository review: $0.52 vs $0.0065. Cache-heavy agent loop: $0.6 vs $0.009.
Context tradeoff
Both models list 1M.
Run the same representative tasks against both endpoints before changing production traffic.

Questions

Which is better, Claude Fable 5.2 or Claude Haiku 5.5?

Claude Fable 5.2 has the higher public point estimate, 91.19 versus 66.12. Their conditional score ranges do not overlap. These ranges do not establish rank confidence. The page therefore keeps the decision tied to the specific documented workload.

Which is better for coding, Claude Fable 5.2 or Claude Haiku 5.5?

Claude Fable 5.2 scores higher for coding on the public lane, 86.9 to 51.7. Claude Fable 5.2 is scored on Estimated evidence for coding, so the reading is directional rather than like-for-like. BenchLM does not name a winner for coding from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which is better for agentic tasks, Claude Fable 5.2 or Claude Haiku 5.5?

Claude Fable 5.2 scores higher for agentic tasks on the public lane, 90.7 to 60.7. Claude Fable 5.2 is scored on Estimated evidence for agentic tasks, so the reading is directional rather than like-for-like. BenchLM does not name a winner for agentic tasks from a directional reading; read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Claude Fable 5.2 or Claude Haiku 5.5?

For the stated presets, chat costs $0.028 on Claude Fable 5.2 and $0.00035 on Claude Haiku 5.5; repository review costs $0.52 and $0.0065; the cache-heavy agent loop costs $0.6 and $0.009. Costs use the listed standard API rates.

Which has the larger context window, Claude Fable 5.2 or Claude Haiku 5.5?

Both models list the same context window, 1M.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence43 rows

Agentic

  • Terminal-Bench 4.0

    Claude Fable 5.267.60%
    Source
    Claude Haiku 5.539.20%
    Source

    Claude Fable 5.2 leads this result

  • Terminal-Bench-Science 0.1

    Claude Fable 5.274.6%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • AutomationBench

    Claude Fable 5.251.4%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • OSWorld 2.1 Offline

    Claude Fable 5.289.3%
    Source
    Claude Haiku 5.572.4%
    Source

    Claude Fable 5.2 leads this result

  • WANDR

    Claude Fable 5.282.1%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • Remote Labor Index

    Claude Fable 5.241.3%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • RoboDojo Sim

    Claude Fable 5.264.9%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • DRACO

    Claude Fable 5.287.7%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • Agents' Last Exam (Linux)

    Claude Fable 5.2—
    Claude Haiku 5.557.0%
    Source

    Not directly comparable

  • HLE w/ tools

    Claude Fable 5.2—
    Claude Haiku 5.557.4%
    Source

    Not directly comparable

  • MazeBench

    Claude Fable 5.2—
    Claude Haiku 5.51%
    Source

    Not directly comparable

  • Hermes Index

    Claude Fable 5.2—
    Claude Haiku 5.537.0%
    Source

    Not directly comparable

  • Hermes Bench

    Claude Fable 5.2—
    Claude Haiku 5.571.0%
    Source

    Not directly comparable

Coding

  • SWE-bench Pro

    Claude Fable 5.293%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • SWE Multilingual

    Claude Fable 5.296.6%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • SWE Multimodal

    Claude Fable 5.274.1%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • DeepSWE

    Claude Fable 5.274.3%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • FrontierSWE v2

    Shared source
    Claude Fable 5.269.9%
    Claude Haiku 5.543.8%

    Claude Fable 5.2 leads this result

  • ProgramBench

    Claude Fable 5.285.6%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • FrontierCode 2.0

    Shared source
    Claude Fable 5.249.1%
    Claude Haiku 5.527.7%

    Claude Fable 5.2 leads this result

  • CursorBench 4.0

    Claude Fable 5.260.0%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • FrontierCode 1.1 Main

    Claude Fable 5.2—
    Claude Haiku 5.546.4%
    Source

    Not directly comparable

  • Bug Hunt Bench

    Claude Fable 5.2—
    Claude Haiku 5.521.5 fixes
    Source

    Not directly comparable

Reasoning

  • ARC-AGI-3

    Claude Fable 5.265.7%
    Source
    Claude Haiku 5.50.5%
    Source

    Claude Fable 5.2 leads this result

  • ARC-AGI-3 Provider Adapter

    Claude Fable 5.294.90%
    Source
    Claude Haiku 5.52.82%
    Source

    Claude Fable 5.2 leads this result

  • STEM Professor Bench

    Claude Fable 5.261.8%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • ARC-AGI-1

    Claude Fable 5.2—
    Claude Haiku 5.589.00%
    Source

    Not directly comparable

  • ARC-AGI-2

    Claude Fable 5.2—
    Claude Haiku 5.561.3%
    Source

    Not directly comparable

Multimodal

  • Chartography (no tools)

    Claude Fable 5.273.4%
    Source
    Claude Haiku 5.546.4%
    Source

    Claude Fable 5.2 leads this result

  • Chartography (tools)

    Claude Fable 5.291.6%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • OfficeQA

    Claude Fable 5.283.4%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • OfficeQA Pro

    Claude Fable 5.272.9%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • BenchCAD 2.1

    Claude Fable 5.275.4%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • BenchCAD 2.1 (tools)

    Claude Fable 5.283.5%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • PerceptionBench (tools)

    Claude Fable 5.2—
    Claude Haiku 5.578.6%
    Source

    Not directly comparable

  • ParseBench

    Claude Fable 5.2—
    Claude Haiku 5.556.02%
    Source

    Not directly comparable

Knowledge

  • HLE

    Claude Fable 5.274.2%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • HLE w/o tools

    Claude Fable 5.273.2%
    Source
    Claude Haiku 5.545.9%
    Source

    Claude Fable 5.2 leads this result

  • PhysicianBench

    Claude Fable 5.264.4%
    Source
    Claude Haiku 5.5—

    Not directly comparable

Multilingual

  • GMMLU

    Claude Fable 5.296.0%
    Source
    Claude Haiku 5.5—

    Not directly comparable

  • MILU

    Claude Fable 5.294.1%
    Source
    Claude Haiku 5.5—

    Not directly comparable

Math

  • FrontierMath v2 (Tiers 1-3)

    Claude Fable 5.2—
    Claude Haiku 5.575.088%
    Source

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Claude Fable 5.2—
    Claude Haiku 5.546.341%
    Source

    Not directly comparable

43 public results · 8 shared

Related comparisons

Compare API pricingRead the methodologyOpen the model selector

Watch Claude Fable 5.2 vs Claude Haiku 5.5

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 6,100+ subscribers.

Last updated November 9, 2026