Compare Models

Mercury 2.5

Mercury 2.5 is a model from Inception, released September 8, 2026. It ranks #44 of 44 models on the Vals Index with 8.86%. Its best result is #59 of 149 on LegalBench.

Release Date: Sep 8, 2026

Developer Inception  🇺🇸
Context Window 260k
Max Output Tokens 66k
Token Costs (in/out) $0.20/0.75
Weights Private
Input Modalities

Accuracy

8.86 % ± 0.45

Cost / Test (Vals Index)

$ 0.286

Latency

5 min 42 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

8.86%

±0.45
44/44
Code Migration

4.52%

±1.05
70/74
CyberBench v1.1

45.48%

±6.06
37/44
EMB

8.79%

±1.27
70/71
Finance Agent (v2)

18.89%

±0.22
72/75
Legal Research Bench

4.33%

±1.41
69/74
MedCode

31.33%

±1.95
93/105
MedScribe

55.09%

±2.10
104/107
ProofBench v1.1

3.00%

±1.71
45/47
Public Benefits Bench v1.1

45.26%

±1.29
44/47
Tax Agent Bench

12.41%

±1.80
66/66
Vibe Code Bench v1.1

3.68%

±1.41
98/109
Harvey's Legal Agent Benchmark

0.00%

±0.00
60/75
IOI

2.44%

±0.68
41/41
LegalBench

83.06%

±0.48
59/149
SkillsBench

18.06%

±3.08
35/35
Terminal-Bench 4.0

0.00%

±0.00
44/44
Terminal-Bench Science

0.00%

±0.00
35/38
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Inception
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 65,536
Reasoning Effort: high