bench.turing BENCHMARK FAMILY
Methodology Changelog
auto_awesomeTURING
Standardized AI-agent evaluation

We benchmark real
economic work.

A growing catalogue across six sectors, one template. Agents are scored on multi-hour operational tasks drawn from private, never-indexed enterprise data — validated against human-expert baselines, with uncertainty and methodology exposed at every level.

HUB LAYOUT
{{ variantSub }}
{{ s.value }}{{ s.unit }}
{{ s.label }}
verifiedHUMAN-EXPERT VALIDATED
shieldCONTAMINATION-CHECKED
lockPRIVATE · NEVER-INDEXED DATA
search
{{ filter.countText }} insights{{ filter.focusHelper }}
{{ b.icon }}
{{ b.name }}{{ b.status }}
{{ b.domain }}

{{ b.positioning }}

{{ c.k }} {{ c.v }}
{{ focusModelInitial }}
{{ focusModelName }} on this bench
reward {{ b.focusRewardLabel }} · {{ b.focusRankLabel }}
not yet evaluated on this bench
{{ b.focusDeltaLabel }}
{{ b.leaderInitial }}
CURRENT LEADER
{{ b.leader }} {{ b.leaderScore }}
 {{ b.noLeaderText }}
search_off
No benches match your filters

Model performance across the family

One model, judged across the enterprise. Pick the benches and models to compare — each axis is soft-gated reward (0–0.60).

BENCHES
{{ ax.name }}
MODELS
{{ l.name }}

Latest updates

ACROSS THE FAMILY
{{ u.tag }}
{{ u.text }}
{{ u.bench }} · {{ u.date }}
The full bench family, one row each. Click any column to sort; click a row to enter that bench. Scales to the whole catalogue.
imagedata_objectlink
{{ r.icon }}
{{ r.name }}
{{ r.domain }}
{{ r.category }} {{ r.tasksLabel }} {{ r.hoursLabel }} {{ r.modelsLabel }}
{{ r.leaderInitial }} {{ r.leader }} —
{{ r.rewardLabel }} {{ r.status }} chevron_right
search_off
No benches match your filters
live preliminary planned Reward = top model's soft-gated score, gated ×0.3 on any core failure.
A map of knowledge work. Each region is a sector of the economy; lit regions are benchmarked today and link to their bench, dimmed regions are planned.
{{ r.icon }} PLANNED north_east
{{ r.name }}
{{ r.sub }}
benchmarked today planned 4 of 8 sectors live · coverage expanding quarterly