Dolphin-v2: benchmark results

Provider: Cognitive Computations.

Aggregate Rating 1702 ± 51, rank #321 of 10552 rated models, from 32 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
MPDocBench-Parse - Heading Hierarchy50.04Tree-edit-distance similarity (0-100; predicted Markdown hea94.4
MPDocBench-Parse - Table Recognition83.84TEDS (0-100; recognized HTML tables)83.3
MPDocBench-Parse - Truncated Table Merging64.28TEDS (0-100; tables split across pages, matched to the merge83.3
OmniDocBench 1.588.71Overall (mixed sources)55.7
OmniDocBench v1.6 Full - Reading Order Edit0.15Normalized edit distance45.7
MPDocBench-Parse73.14Overall score (0-100; mean of the eight task scores, with th44.4
OmniDocBench v1.6 Full - Table TEDS84.4Table TEDS (%)42.9
OmniDocBench v1.6 Full - Formula CDM91.01Formula CDM (%)40
PureDocBench - Clean - Formula CDM59.8Formula Character Detection Matching score (0-100) between t36.8
PureDocBench - Clean - Reading Order Edit Distance42.9Reading-order normalized edit distance (%) between the predi36.8
OmniDocBench v1.6 Full - Overall89.5Overall (%)34.3
PureDocBench - Clean - Table TEDS72.12Table Tree-Edit-Distance-based Similarity (0-100) between th33.3

Interactive version: theaggregate.ai/model?slug=dolphin-v2 · How It Works · Data refreshed daily, snapshot 2026-11-16.