Legal task processing on Harvey LAB benchmark leaderboard · SOTA2 Research

Legal task processing on Harvey LAB

11.99All-Pass RatePARTHENON

Evaluation Results

Method	Links
PARTHENON2026.06		11.99	90.2	89.9	90.1	92.4	90.3	90.6	86.5	88.6	88.4	84.1	85.9	91.4	91.5	89.9	89.4	91.2	90.6	92.9	90.7	89.5	94.8	86.5	89.1	91.5	93.8
Claude Code2026.06		11.83	82.8	76.5	74.5	76.5	80.2	86.4	75.2	82	82.4	83.9	79	80.9	85.7	85.2	81.6	82.1	81.9	87.2	86.8	87	86.4	87.3	83.8	87	90.8
PARTHENON2026.06		10.95	89.9	89.6	90.9	89.7	91.1	91.2	86.1	89.9	87.7	87	87.8	92.2	91.8	88.1	89.6	89.6	89.1	92.2	89.5	89.2	90.3	86.3	89.3	91.3	92.8
Codex2026.06		3.76	79.8	69.8	72.3	74.6	75.4	76	77.8	78.3	78.7	78.9	79.5	79.5	80.6	80.7	80.8	81.4	81.5	81.6	82	82.8	83.2	83.3	83.7	84.6	88.6
PARTHENON2026.06		3.36	82	82.3	84.8	82	82.9	81.1	77.7	80.8	77.7	79.8	81.7	82.9	84.1	78	79.7	84.3	81.9	84.8	81.1	81.8	85.4	78.8	80.5	83	85.6
PARTHENON2026.06		3.04	77.8	79.9	80	81.3	79.7	75.1	79.6	74.2	75.4	69.8	80.6	81.9	79.2	75.7	70.8	76.6	78.9	85.5	79.7	76.3	80.9	68.1	75	75.3	76.9
Codex2026.06		1.12	68.2	56.7	65.1	65.2	66.2	68.9	63.1	68.8	64.4	68.7	67.4	66.2	68.9	69	70.2	64	71.4	69.3	70.7	68.3	72.6	73.5	72.8	73.3	64.2
Claude Code2026.06		1.12	71.6	66.3	67.6	71.7	70.5	67.7	59.5	64.8	68.6	66.4	72.7	70.9	73.9	72.3	70.4	69.3	71.2	77.4	74.6	72.1	83.9	63.9	73.5	74.8	75.3
Direct Prompting (API)2026.06		0.4	56.5	47.4	44.1	55.1	63.3	54.2	57.2	44.4	58	50.4	59.9	61.2	61.4	62.4	53.1	58	58.3	66.9	60.6	61	70	51.4	52.1	60.5	58.8
Basic Legal-Native Harness2026.06		0.16	60.7	61	67.9	62.8	62.8	62.9	54.2	64.2	59.8	48	54.7	60.8	64.3	56.2	59.4	59.6	53.6	65.4	54.8	52.2	63.1	54.1	63.2	56.3	63.3