PRO Bench / Public evidence

Compare systems.Inspect the proof.

Questions, participant identities, results, cost, time and evaluator evidence belong in one inspectable chain. Missing evidence stays unknown.

Publication status JSON Read methodology

Current public state

No benchmark result is public until an immutable evidence package passes every gate.

Observed application build: da50cacfac4f

Evidence target

A rights-cleared, integrity-verified R3 snapshot with exact questions, outputs and receipts.

No verified R3 evidence snapshot is configured for public serving.

Evidence architecture

From task to proof.

Every published result must resolve through the same inspectable chain.

Inspect the rules

01Challenge

Exact task, version and permitted data.

02Participant

System class, build and tool policy.

03Attempt

Output, artifacts, cost and timing.

04Evaluation

Rubric, judges and limitations.

05Publication

Rights-cleared immutable package.

Benchmark catalog

Open a suite to inspect its scope, proof state and per-item evidence boundaries.

No public suites yet.Catalog data is empty; no score is inferred.

Participant tracks

Compare like with like

Raw models, chat products, coding agents, agent runtimes and orchestrators are different participants. Official ranks stay inside a frozen comparison protocol.

Planned coverage →

comparison trackModel onlyMODEL_ONLY

One pinned model snapshot, no external tools or product workflow.

pinned_model_snapshot
frozen_prompt_protocol
no_external_tools

comparison trackModel + standard toolsSTANDARD_TOOLS

Pinned model snapshots receive the same frozen browser, code and file tools.

pinned_model_snapshot
identical_tool_contract
frozen_prompt_protocol

comparison trackAgent / systemAGENT_SYSTEM

Complete products, coding agents, agent runtimes and orchestrators compete as systems.

product_or_system_snapshot
interaction_protocol
tool_and_model_disclosure

comparison trackBest result under equal budgetEQUAL_BUDGET

Different participant classes may compete only under the same money, time and operator envelope.

shared_money_cap
shared_wall_clock_cap
shared_operator_protocol
actual_cost_and_time_receipts

Live execution

Execution progress is not qualification. Build identity and evidence remain separate.

Open live view →

questions 0

completed runs 0/0

updated UNKNOWN

No active public execution.Live totals remain UNKNOWN until a run is bound.

Machine-readable data

The interface is a view. These endpoints expose the same fail-closed state for verification.

Catalog JSON Live JSON Participant tracks JSON Publication JSON Route manifest