Human reference · human tier · October 2026 (demo)
Demo data · not real resultsHuman baseline
Navigator Index88.095% CI 84.5–91.2
RankUnrankedReference row: shown, never rankedunranked
Cost / successful task$0.97No task passed, so no cost per success
Median latency4.0 minPass rate 92% · every time 87%
01Categories
Where it is strong, and where it is not
Category · weightHuman baseline
Coding 15%83.373.1–93.6 · 5 tasks
Data analysis 15%90.887.1–94.6 · 5 tasks
Extraction 10%82.559.2–97.4 · 5 tasks
Reasoning 15%86.779.2–93.2 · 5 tasks
Research & agentic 15%For now these are single-call multi-step tasks: one model call returns plan → steps → deliverable, checked step by step. Tool-using, browsing and long agentic tasks are deferred.93.389.6–98.0 · 5 tasks
Summarization 10%85.075.7–93.4 · 5 tasks
Support 10%85.073.8–96.2 · 5 tasks
Writing 10%95.891.2–99.6 · 5 tasks
Run receipts120 demo runs · no receipts
02Cost
What a working answer costs
$0.97per successful task
$965per 1,000 successes
$96,525per 100,000 successes
92%runs that pass
120runs measured
0 of 13models it dominates
Cost per successful task divides everything measured on candidate calls by the tasks that passed, so failed runs are priced in. Judge cost is excluded. Definition
Dominates
No model: none is both significantly worse on the Index and dearer per successful task.
Dominated by
Nothing on this board beats it on both.
03Trend
Across published cycles
Navigator Index — higher is better · shaded: 95% interval
Tracking since 14 Oct 2026
Published cycle
04Receipts
Every run, on the record
Demo data · not real results This is demo data: these runs do not exist, so there are no receipts. On a published cycle this lists every run with its prompt, output, checks and measured cost.
Navigator Bench · Index v2.0 · data through 14 Oct 2026Methodology