Skip to content
AI Atlas
CompareModel

Claude Fable 5.1 vs Claude Sonnet 5 vs Claude Haiku 4.5

3 models · 24 recorded dimensions. Each cell is the value as stated by its source; benchmarks only where every entity sits in the same comparability group.

Tray 3 / 6

  • ModelClaude Fable 5.1Anthropic
  • ModelClaude Sonnet 5Anthropic
  • ModelClaude Haiku 4.5Anthropic
Compare 3 →
Differences only
Emphasise
Comparison of 3 entities
DimensionClaude Fable 5.1AnthropicClaude Sonnet 5AnthropicClaude Haiku 4.5Anthropic
Overview
Release date
Context
Context windowtokens
Max outputtokens
Knowledge cutoff
Benchmarks
Artificial Analysis Intelligence IndexindexPartialreasoning_effort: max vs unspecified; reasoning: unspecified vs onIndependent
GPQA Diamondaccuracy · variant=Diamond · evaluator=Artificial AnalysisPartialreasoning_effort: max vs unspecified; reasoning: unspecified vs onIndependent
GPQA Diamondaccuracy · variant=GPQA Diamond · evaluator=Artificial AnalysisPartialreasoning: unspecified vs onIndependent
Humanity's Last Examaccuracy · evaluator=Artificial AnalysisPartialreasoning_effort: max vs unspecified; reasoning: unspecified vs onIndependent
SciCodeaccuracy · evaluator=Artificial AnalysisPartialreasoning_effort: max vs unspecified; reasoning: unspecified vs onIndependent
Terminal-Benchaccuracy · variant=v2.1 · evaluator=Artificial AnalysisPartialreasoning_effort: max vs unspecified; reasoning: unspecified vs onIndependent
Terminal-Benchaccuracy · variant=v4.0 · evaluator=Artificial AnalysisPartialreasoning_effort: xhigh vs max; reasoning_effort: xhigh vs unspecified; reasoning: unspecified vs onIndependent
Pricing
Best input priceUSD / 1M tokens
Best output priceUSD / 1M tokens
History
First seen in AI Atlas
11 Sept 2026
11 Sept 2026
11 Sept 2026
Last change
12 Sept 2026
12 Sept 2026
12 Sept 2026
Pricing · cheapest deployment per provider (USD / 1M in / out)
Anthropic API$5 / $25$1 / $5$0.50 / $2.5
OpenRouter$5 / $25$1 / $5$0.50 / $2.5
Hardware fit
Estimated memoryestimatedNo parameter count — not estimableNo parameter count — not estimableNo parameter count — not estimable

15 rows shown, 11 identical hidden. Bold = best number in a row (highest; lowest for prices) — a reading aid, not a verdict. Benchmarks appear only when every entity has a current result in the same comparability group. Benchmark dimensions appear only when every compared model has a current result in the same comparability group (benchmark × metric × config_key). Click any value for its evidence.

Share this comparison with its URL. Values are the atlas as of now; use each entity's change history for how a value moved. How AI Atlas records facts →

Claude Fable 5.1 vs Claude Sonnet 5 vs Claude Haiku 4.5 — model comparison | AI Atlas