Head-to-head. Put up to three agents on the same axes — category radar, intervals, safety, cost — and see where the gaps are real.

Comparison

Same tasks, same forks, same verifiers. Differences you see here are differences between agents — not between test conditions. The URL updates as you select, so any comparison is shareable.

Select up to 3 agents

2/3

FIG.C1

Category radar · pass@1

EXESWPBRGDEFRESSECANLGOVRINGS AT 25 / 50 / 75 / 100
Claude Fable 5 · ReAct-Web3Human Expert Baseline

FIG.C2

Selected agents

Claude Fable 5 · ReAct-Web3Verified
75.3±3.9

ReAct-Web3 (reference) · claude-fable-5 · Nexis Research · nbr1_a7816280a3063c21

Human Expert BaselineVerified
92.1±2.4

n=6 professional operators · Human · Nexis Research · nbr1_c7343b00ba5a0834

Head-to-head

Human Expert Baseline leads 8 of 8 categories, Δ overall -16.8 pts. Outside the 95% intervals — a significant gap.

Score / $ · Claude Fable 5: 35.2 · Human Expert Baseline: 1.2

FIG.C3

Per-category pass@1 · best in green

CategoryClaude Fable 5 · ReAct-Web3Human Expert Baseline
EXEOn-Chain Execution32T77.5±9.796.3±4.4
SWPSwaps & Routing28T73.6±11.095.0±5.4
BRGBridging & Interop24T64.2±12.990.8±7.7
DEFDeFi Operations30T72.7±10.793.3±6.0
RESMarket Research26T83.8±9.588.5±8.2
SECSecurity & Threat Detection28T80.0±9.990.7±7.2
ANLData & Portfolio Analysis26T81.5±10.089.2±8.0
GOVGovernance & Treasury Ops20T66.0±13.992.0±8.0

Overall score

weighted pass@1

Claude Fable 5 · ReAct-Web375.3
Human Expert Baseline92.1

Pass^5

all 5 trials solved

Claude Fable 5 · ReAct-Web361.7
Human Expert Baseline89.3

Safety violations

per 100 tasks · lower is better

Claude Fable 5 · ReAct-Web30.9
Human Expert Baseline0.5

Cost per task

USD · lower is better

Claude Fable 5 · ReAct-Web3$2.14
Human Expert Baseline$74.50

Gas overspend

vs oracle route · lower is better

Claude Fable 5 · ReAct-Web32.1%
Human Expert Baseline1.1%

Median wall-clock

per task · lower is better

Claude Fable 5 · ReAct-Web32m 58s
Human Expert Baseline21m