SShogo Bench

Claude Fable 5.2 vs Claude Haiku 5.5

Claude Fable 5.2 leads Claude Haiku 5.5 by 18.9 points on the Shogo Agent Index. Claude Fable 5.2 wins 19 and Claude Haiku 5.5 wins 0 of 19 shared benchmarks. Claude Haiku 5.5 is cheaper for agent workloads.

Add more models to this comparison →

Pillars

PillarClaude Fable 5.2Claude Haiku 5.5
Agentic coding8667
Long-horizon autonomy7862
Knowledge-work agents6746
Computer use7747
Security & context learning6555

Shared benchmarks

BenchmarkClaude Fable 5.2Claude Haiku 5.5
APEX-Agentsindex80.1%56.1%
Vending-Bench 2index$9,628$3,602
FrontierSWEindex69.9%43.8%
CursorBenchindex60.0%48.4%
ARC-AGI98.5%89.0%
ARC-AGI-293.8%61.3%
Chess Puzzles79.0%55.0%
CritPt35.7%22.6%
DTBench99.2%95.0%
FrontierMath Tier 4100.0%46.3%
FrontierMath Tiers 1-395.1%75.1%
GPQA Diamond96.0%89.6%
LMCA70.6%51.1%
OTIS Mock AIME99.7%98.9%
ProofBench100.0%89.0%
SciCode67.7%55.0%
SimpleBench94.9%59.2%
SimpleQA Verified94.0%23.8%
WeirdML v343.1%11.7%

Specs and price

Claude Fable 5.2Claude Haiku 5.5
Agent Index76.4 (#1)57.5 (#18)
Agent $/M$3.21$0.050
Input / output $/M$8.00 / $40.00$0.10 / $0.50
Context1M1M
ReleasedNov 9, 2026Oct 7, 2026
WeightsClosedClosed
API hosts45

Don’t pick one. Route.

Shogo can run Claude Fable 5.2 and Claude Haiku 5.5 side by side and send each step of an agent to the one that’s cheaper for it.

Try both in Shogo