Agents’ Last ExamBrowse 296

Agents’ Last Exam

Can the model complete long-running professional workflows across many occupations?

Living benchmarkLiving public release
Knowledge Work28 ranked models78 reported values2 reportsHigher is better

Top rankings

One row per model, using its best reported score across effort settings.

28
RankModelBest scoreBest reported setting
1GPT-6 AstraOpenAI59.3%max effortest. $6.23 / task6 values · 1 reportSep 22, 2026 · source
2Claude Opus 5Anthropic55.9%high effortest. $7.29 / task5 values · 1 reportSep 22, 2026 · source
3Muse Spark 1.3Meta55.8%xhigh effortest. $2.82 / task2 values · 1 reportSep 22, 2026 · source
4GPT-5.6 SolOpenAI53.6%xhigh effortest. $5.08 / task6 values · 2 reportsSep 22, 2026 · source
5Qwen3.8 MaxQwen52.5%xhigh effortest. $3.20 / task1 value · 1 reportSep 22, 2026 · source
6Kimi K3Moonshot AI51.6%max effortest. $3.99 / task2 values · 1 reportSep 22, 2026 · source
7Grok 4.5SpaceXAI51.2%high effortest. $2.22 / task1 value · 1 reportSep 22, 2026 · source
8GPT-5.6 TerraOpenAI50.7%max effortest. $3.58 / task6 values · 2 reportsSep 22, 2026 · source
9GPT-5.6 LunaOpenAI50.4%max effortest. $2.57 / task6 values · 2 reportsSep 22, 2026 · source
10Claude Fable 5Anthropic48.7%xhigh effortest. $28.55 / task3 values · 2 reportsSep 22, 2026 · source
11GPT-5.5OpenAI47.9%xhigh effortest. $3.96 / task9 values · 2 reportsSep 22, 2026 · source
12Claude Opus 4.8Anthropic45.2%Reported configurationOpenAI report5 values · 2 reportsJul 9, 2026 · source

Effort curve

Every sourced cost-linked effort value for this exact version. Lines connect complete sweeps only.

70
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · claude_code · published cost basisClaude Opus 4.8

+4.5est. cost · 2.5× cost

EffortScoreCost/task
max45.1%est. $26.22
xhigh42.3%est. $18.52
high40.7%est. $10.70
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · claude_code · official cost basisClaude Opus 5

+0.8est. cost · 2.4× cost

EffortScoreCost/task
max52.7%est. $9.76
xhigh55.5%est. $10.02
high55.9%est. $7.29
medium53.0%est. $5.29
low51.9%est. $4.06
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · codex · db cost basisGPT-5.5

+11.6est. cost · 4.1× cost

EffortScoreCost/task
xhigh47.9%est. $3.96
medium39.8%est. $2.35
low36.4%est. $0.96
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · codex · list cost basisGPT-5.6 Luna

+19.3est. cost · 14.6× cost

EffortScoreCost/task
max50.4%est. $2.57
xhigh49.4%est. $1.55
high46.1%est. $0.84
medium38.5%est. $0.38
low31.0%est. $0.18
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · codex · db+list cost basisGPT-5.6 Sol

+7.7est. cost · 4.4× cost

EffortScoreCost/task
max52.8%est. $7.13
xhigh53.6%est. $5.08
high52.4%est. $3.79
medium52.1%est. $3.41
low45.1%est. $1.63
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · codex · db+list cost basisGPT-5.6 Terra

+6.3est. cost · 2.9× cost

EffortScoreCost/task
high46.6%est. $1.72
medium42.6%est. $0.84
low40.2%est. $0.59
UC Berkeley RDI / Agents' Last Exam · ALE API snapshot · codex · official cost basisGPT-6 Astra

+5.9est. cost · 2.0× cost

EffortScoreCost/task
max59.3%est. $6.23
xhigh58.3%est. $5.40
high57.8%est. $4.64
medium57.6%est. $4.10
low53.4%est. $3.07
Definition and comparison boundarypublic methodology

Can the model complete long-running professional workflows across many occupations?

Average partial-credit performance on end-to-end agent work across 55 professional subdomains in reproducible desktop sandboxes. Higher is better. Average partial-credit score expressed as a percentage.

ALE-V1 is a living benchmark. Model, harness, effort variant, run count, and task coverage all matter; do not compare this average score with ALE's perfect-run pass rate as if they were the same measurement.

ALE-V1 full/overall API snapshot; harness claude_code; harness variant variant not reported; 152 runs covering 144 of 152 tasks; 35 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $2314.73 with raw costSource official. This configuration is shown as an unconnected comparison point at $15.228486842105 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant reasoning-xhigh; 150 runs covering 150 of 152 tasks; 39 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $4340.24 with raw costSource official. This configuration is shown as an unconnected comparison point at $28.554210526316 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant variant not reported; 146 runs covering 143 of 152 tasks; 30 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $935.7 with raw costSource db. This configuration is shown as an unconnected comparison point at $6.155921052632 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw_cli; harness variant variant not reported; 144 runs covering 144 of 152 tasks; 19 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $439.63 with raw costSource db. This configuration is shown as an unconnected comparison point at $2.892302631579 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness ale_claw; harness variant variant not reported; 143 runs covering 143 of 152 tasks; 28 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1132.24 with raw costSource official. This configuration is shown as an unconnected comparison point at $7.448947368421 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant variant not reported; 137 runs covering 137 of 152 tasks; 21 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1792.72 with raw costSource official+adj. This configuration is shown as an unconnected comparison point at $11.794210526316 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness cursor_cli; harness variant thinking-high; 137 runs covering 137 of 152 tasks; 32 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1899.32 with raw costSource official+adj. This configuration is shown as an unconnected comparison point at $12.495526315789 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness droid; harness variant variant not reported; 105 runs covering 99 of 152 tasks; 22 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1356.25 with raw costSource official+adj. This configuration is shown as an unconnected comparison point at $8.922697368421 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 142 runs covering 141 of 152 tasks; 24 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1688.85 with raw costSource official+adj. This configuration is shown as an unconnected comparison point at $11.110855263158 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant variant not reported; 141 runs covering 141 of 152 tasks; 25 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1837.79 with raw costSource official+adj. This configuration is shown as an unconnected comparison point at $12.090723684211 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-high; 143 runs covering 143 of 152 tasks; 34 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1626 with raw costSource published. This eligible same-harness effort sweep normalizes that split total to $10.697368421053 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-max; 150 runs covering 150 of 152 tasks; 41 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $3985 with raw costSource published. This eligible same-harness effort sweep normalizes that split total to $26.217105263158 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-xhigh; 151 runs covering 151 of 152 tasks; 34 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $2815 with raw costSource published. This eligible same-harness effort sweep normalizes that split total to $18.519736842105 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-high; 150 runs covering 150 of 152 tasks; 49 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1108.34 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $7.291710526316 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-low; 150 runs covering 150 of 152 tasks; 42 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $617.41 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $4.061907894737 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-max; 148 runs covering 148 of 152 tasks; 47 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1483.26 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $9.758289473684 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-medium; 150 runs covering 150 of 152 tasks; 44 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $804.19 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $5.290723684211 estimated per task. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-xhigh; 151 runs covering 151 of 152 tasks; 46 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1523.43 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $10.022565789474 estimated per task. ALE-V1 full/overall API snapshot; harness cursor_cli; harness variant variant not reported; 138 runs covering 138 of 152 tasks; 31 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $176.65 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $1.162171052632 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 275 runs covering 146 of 152 tasks; 39 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $273.05 with raw costSource db. This configuration is shown as an unconnected comparison point at $1.796381578947 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness gemini_cli; harness variant variant not reported; 142 runs covering 142 of 152 tasks; 25 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $2018.05 with raw costSource db+list+adj. This configuration is shown as an unconnected comparison point at $13.276644736842 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 149 runs covering 141 of 152 tasks; 22 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $2968.62 with raw costSource db+adj. This configuration is shown as an unconnected comparison point at $19.530394736842 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 260 runs covering 140 of 152 tasks; 35 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $439.78 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $2.893289473684 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant max; 146 runs covering 146 of 152 tasks; 31 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1085.62 with raw costSource db. This configuration is shown as an unconnected comparison point at $7.142236842105 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness ale_claw; harness variant variant not reported; 143 runs covering 143 of 152 tasks; 18 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $335.05 with raw costSource db. This configuration is shown as an unconnected comparison point at $2.204276315789 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant variant not reported; 144 runs covering 142 of 152 tasks; 11 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $134.93 with raw costSource db. This configuration is shown as an unconnected comparison point at $0.887697368421 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 179 runs covering 142 of 152 tasks; 40 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $274.1 with raw costSource db. This configuration is shown as an unconnected comparison point at $1.803289473684 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness ale_claw; harness variant variant not reported; 145 runs covering 145 of 152 tasks; 35 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $309.77 with raw costSource db. This configuration is shown as an unconnected comparison point at $2.037960526316 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant variant not reported; 438 runs covering 152 of 152 tasks; 110 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $415.59 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $2.734144736842 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-low; 152 runs covering 152 of 152 tasks; 26 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $146.61 with raw costSource db. This eligible same-harness effort sweep normalizes that split total to $0.964539473684 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-medium; 149 runs covering 149 of 152 tasks; 28 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $357.68 with raw costSource db. This eligible same-harness effort sweep normalizes that split total to $2.353157894737 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 299 runs covering 152 of 152 tasks; 80 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $601.69 with raw costSource db. This eligible same-harness effort sweep normalizes that split total to $3.958486842105 estimated per task. ALE-V1 full/overall API snapshot; harness cursor_cli; harness variant variant not reported; 142 runs covering 139 of 152 tasks; 33 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $173.52 with raw costSource list. This configuration is shown as an unconnected comparison point at $1.141578947368 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness droid; harness variant variant not reported; 139 runs covering 137 of 152 tasks; 30 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $242.24 with raw costSource list. This configuration is shown as an unconnected comparison point at $1.593684210526 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 136 runs covering 136 of 152 tasks; 33 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $446.52 with raw costSource db. This configuration is shown as an unconnected comparison point at $2.937631578947 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-high; 152 runs covering 152 of 152 tasks; 37 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $127.52 with raw costSource list. This eligible same-harness effort sweep normalizes that split total to $0.838947368421 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-low; 152 runs covering 152 of 152 tasks; 18 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $26.77 with raw costSource list. This eligible same-harness effort sweep normalizes that split total to $0.176118421053 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-max; 149 runs covering 149 of 152 tasks; 44 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $390.61 with raw costSource list. This eligible same-harness effort sweep normalizes that split total to $2.569802631579 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-medium; 152 runs covering 152 of 152 tasks; 26 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $57 with raw costSource list. This eligible same-harness effort sweep normalizes that split total to $0.375 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 152 runs covering 152 of 152 tasks; 46 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $235.12 with raw costSource list. This eligible same-harness effort sweep normalizes that split total to $1.546842105263 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-high; 299 runs covering 152 of 152 tasks; 91 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $575.68 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $3.787368421053 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-low; 299 runs covering 151 of 152 tasks; 73 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $248.12 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $1.632368421053 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-max; 307 runs covering 152 of 152 tasks; 89 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $1084.09 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $7.132171052632 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-medium; 302 runs covering 152 of 152 tasks; 87 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $518.72 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $3.412631578947 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 301 runs covering 152 of 152 tasks; 93 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $771.62 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $5.076447368421 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-high; 298 runs covering 151 of 152 tasks; 77 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $261.53 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $1.720592105263 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-low; 300 runs covering 152 of 152 tasks; 60 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $89.49 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $0.58875 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-max; 297 runs covering 152 of 152 tasks; 82 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $544.61 with raw costSource list. This configuration is shown as an unconnected comparison point at $3.582960526316 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-medium; 297 runs covering 152 of 152 tasks; 69 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $128.23 with raw costSource db+list. This eligible same-harness effort sweep normalizes that split total to $0.843618421053 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 299 runs covering 152 of 152 tasks; 82 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $380.73 with raw costSource list. This configuration is shown as an unconnected comparison point at $2.504802631579 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-high; 151 runs covering 151 of 152 tasks; 48 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $705.22 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $4.639605263158 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-low; 151 runs covering 151 of 152 tasks; 45 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $467.03 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $3.072565789474 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-max; 151 runs covering 151 of 152 tasks; 52 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $947.52 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $6.233684210526 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-medium; 151 runs covering 151 of 152 tasks; 49 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $623.9 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $4.104605263158 estimated per task. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-none; 151 runs covering 151 of 152 tasks; 43 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $696.97 with raw costSource official. This configuration is shown as an unconnected comparison point at $4.585328947368 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 151 runs covering 151 of 152 tasks; 49 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $820.35 with raw costSource official. This eligible same-harness effort sweep normalizes that split total to $5.397039473684 estimated per task. ALE-V1 full/overall API snapshot; harness grok_cli; harness variant variant not reported; 143 runs covering 140 of 152 tasks; 7 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $174.28 with raw costSource db. This configuration is shown as an unconnected comparison point at $1.146578947368 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness grok_cli; harness variant variant not reported; 149 runs covering 142 of 152 tasks; 11 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $284.6 with raw costSource list. This configuration is shown as an unconnected comparison point at $1.872368421053 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 262 runs covering 145 of 152 tasks; 13 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $80.18 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $0.5275 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness grok_build; harness variant reasoning-high; 151 runs covering 151 of 152 tasks; 41 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $336.79 with raw costSource db. This configuration is shown as an unconnected comparison point at $2.215723684211 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 241 runs covering 144 of 152 tasks; 25 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $123.45 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $0.812171052632 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-max; 149 runs covering 149 of 152 tasks; 41 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $463.87 with raw costSource db. This configuration is shown as an unconnected comparison point at $3.051776315789 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness kimi_code; harness variant thinking-max; 149 runs covering 149 of 152 tasks; 43 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $605.78 with raw costSource db. This configuration is shown as an unconnected comparison point at $3.985394736842 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 261 runs covering 146 of 152 tasks; 25 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $43.34 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $0.285131578947 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 223 runs covering 146 of 152 tasks; 14 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $26.74 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $0.175921052632 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-max; 148 runs covering 148 of 152 tasks; 45 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $408.58 with raw costSource official. This configuration is shown as an unconnected comparison point at $2.688026315789 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness codex; harness variant reasoning-xhigh; 148 runs covering 148 of 152 tasks; 49 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $428.09 with raw costSource official. This configuration is shown as an unconnected comparison point at $2.816381578947 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness claude_code; harness variant reasoning-max; 148 runs covering 148 of 152 tasks; 31 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost and costSource are null in the API row. The API does not report an Estimated Total Cost or costSource for this configuration, so no cost-per-task value is attached. ALE-V1 full/overall API snapshot; harness claude_code; harness variant thinking-xhigh; 148 runs covering 148 of 152 tasks; 41 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $486.47 with raw costSource db. This configuration is shown as an unconnected comparison point at $3.200460526316 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 142 runs covering 142 of 152 tasks; 18 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $663.87 with raw costSource db+list. This configuration is shown as an unconnected comparison point at $4.367565789474 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. ALE-V1 full/overall API snapshot; harness openclaw; harness variant variant not reported; 272 runs covering 146 of 152 tasks; 25 passes. The API's passRate and avgScore fields are retained as separate measurements. Official Estimated Total Cost is $253.85 with raw costSource db. This configuration is shown as an unconnected comparison point at $1.670065789474 estimated per task because its same-harness, same-cost-source group does not contain at least three effort settings. OpenAI reports ALE's average partial-credit score, not its perfect-run pass rate.

Method / source