Legal Research Bench ↗4.33 %
Mercury 2.5 · Legal Research Bench · All-Pass ↗
Reported model / MODEL EVIDENCE
Mercury 2.5
17 benchmarks · 17 observations · 17 sources
Scores at a glance
17 benchmarks · large dot: Mercury 2.5; small dots: closest models where reported
LegalBench ↗83.06 %
Mercury 2.5 · LegalBench · Accuracy ↗
Mercury 2.5 · Harvey's Legal Agent Benchmark v1 · Accuracy ↗
Tax Agent Bench ↗2.49 %
Mercury 2.5 · Tax Agent Bench · Accuracy ↗
Vals Finance Agent v2 ↗18.89 %
Mercury 2.5 · Finance Agent v2 · Accuracy ↗
Mercury 2.5 · Excel Modeling Benchmark v1 · Accuracy ↗
MedCode ↗31.33 %
Mercury 2.5 · MedCode · Accuracy ↗
MedScribe ↗55.09 %
Mercury 2.5 · MedScribe · Accuracy ↗
ProofBench ↗3 %
Mercury 2.5 · ProofBench · Accuracy ↗
Mercury 2.5 (Mini-SWE-agent) · Terminal-Bench-Science · Score ↗
Vibe Code Bench ↗3.68 %
Mercury 2.5 (OpenHands) · Vibe Code Bench · original task set ↗
Mercury 2.5 (Mini-SWE-agent) · Terminal-Bench 4.0 · Score ↗
SkillsBench ↗18.06 %
Mercury 2.5 (OpenHands) · SkillsBench · Score ↗
IOI v1 & IOI v2 ↗2.44 %
Mercury 2.5 · IOI · Accuracy ↗
Code Migration v1.1 ↗4.52 %
Mercury 2.5 · Code Migration v1 · Accuracy ↗
CyberBench v1.1 ↗45.48 %
Mercury 2.5 · CyberBench · Accuracy ↗
Public Benefits Bench v1.1 ↗45.26 %
Mercury 2.5 · Public Benefits Bench · Accuracy ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
Gemini 3.5 Flash Lite →Google
11 of 17 shared benchmark comparisons
13 of 17 shared benchmark comparisons
13 of 17 shared benchmark comparisons
How similarity is measured
Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- Code Migration v1 — evaluation methodology and results ↗
- CyberBench — evaluation methodology and results ↗
- Excel Modeling Benchmark v1 — evaluation methodology and results ↗
- Finance Agent v2 — evaluation methodology and results ↗
- Harvey's Legal Agent Benchmark v1 — evaluation methodology and results ↗
- IOI — evaluation methodology and results ↗
- LegalBench — evaluation methodology and results ↗
- Legal Research Bench — evaluation methodology and results ↗
- MedCode — evaluation methodology and results ↗
- MedScribe — evaluation methodology and results ↗
- ProofBench — evaluation methodology and results ↗
- Public Benefits Bench — evaluation methodology and results ↗
- SkillsBench — evaluation methodology and results ↗
- Tax Agent Bench — evaluation methodology and results ↗
- Terminal-Bench 4.0 — evaluation methodology and results ↗
- Terminal-Bench-Science — evaluation methodology and results ↗
- Vibe Code Bench — evaluation methodology and results ↗
Names used in source tables
Mercury 2.5 · Mercury 2.5 (Mini-SWE-agent) · Mercury 2.5 (OpenHands)