Measured capability
RAG / Grounding
Answers and citations grounded in a bounded document collection.
1 measured exam4 recorded lanesLatest measurement 2026-09-02Published campaign scopeNo category composite
Certified exam
Grounded question answering
4 measured model lanes
Model lane · condition 1 of 2
cli/claude-sonnet
Historical · retired revision1/1010% certified
Lanecli/claude-sonnetDeliveryDelivery not establishedConditionsUnproven
Model failures9Lane errors0Withheld / invalid0 / 0
Elapsed median51.6 seconds · 10/10 observedSteps median11.5 · 10/10 observedTokens medianNot recordedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane · condition 2 of 2
cli/claude-sonnet
Historical · retired revision0/1Certified passes · rate not published
Lanecli/claude-sonnetDeliveryDelivery not establishedConditionsUnproven
Model failures1Lane errors0Withheld / invalid0 / 0
Elapsed median71.8 seconds · 1/1 observedSteps median12 · 1/1 observedTokens medianNot recordedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane · condition 1 of 2
cli/codex
Historical · retired revision10/10100% certified
Lanecli/codexDeliveryDelivery not establishedConditionsUnproven
Model failures0Lane errors0Withheld / invalid0 / 0
Elapsed median66.2 seconds · 10/10 observedSteps median14 · 10/10 observedTokens medianNot recordedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane · condition 2 of 2
cli/codex
Historical · retired revision1/1Certified passes · rate not published
Lanecli/codexDeliveryDelivery not establishedConditionsUnproven
Model failures0Lane errors0Withheld / invalid0 / 0
Elapsed median68.5 seconds · 1/1 observedSteps median14 · 1/1 observedTokens medianNot recordedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane · condition 1 of 2
codestral-latest
Historical · retired revision0/9Certified passes · rate not published
Lanecodestral-latestDeliveryDelivery not establishedConditionsUnproven
Model failures9Lane errors1Withheld / invalid0 / 0
Elapsed median16.5 seconds · 9/9 observedSteps median9 · 9/9 observedTokens median23906 · 9/9 observedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane · condition 2 of 2
codestral-latest
Historical · retired revision0/1Certified passes · rate not published
Lanecodestral-latestDeliveryDelivery not establishedConditionsUnproven
Model failures1Lane errors0Withheld / invalid0 / 0
Elapsed median18.7 seconds · 1/1 observedSteps median10 · 1/1 observedTokens median27288 · 1/1 observedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →
Model lane
openai/gpt-oss-120b
Historical · retired revisionNot evaluatedNo graded attempts
Laneopenai/gpt-oss-120bDeliveryDelivery not establishedConditionsUnproven
Model failures0Lane errors1Withheld / invalid0 / 0
Elapsed medianNot recordedSteps medianNot recordedTokens medianNot recordedCost median (USD)Not recorded
delivery equivalence is unproven; delivery is not proved; lane configuration is not recorded; recorded engine or harness conditions differ
Inspect measurement evidence →