GPT-5.2 vs Mercury 2

Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.

GPT-5.2 is clearly ahead on the aggregate, 88 to 65. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

GPT-5.2's sharpest advantage is in coding, where it averages 81.8 against 41.1. The single biggest benchmark swing on the page is SWE-bench Pro, 85 to 43.

GPT-5.2 is also the more expensive model on tokens at $2.00 input / $8.00 output per 1M tokens, versus $0.25 input / $0.75 output per 1M tokens for Mercury 2. That is roughly 10.7x on output cost alone. GPT-5.2 gives you the larger context window at 400K, compared with 128K for Mercury 2.

Quick Verdict

Pick GPT-5.2 if you want the stronger benchmark profile. Mercury 2 only becomes the better choice if you want the cheaper token bill.

Agentic

GPT-5.2

GPT-5.2

85.4

Mercury 2

63.7

90
Terminal-Bench 2.0
63
84
BrowseComp
67
81
OSWorld-Verified
62

Coding

GPT-5.2

GPT-5.2

81.8

Mercury 2

41.1

91
HumanEval
75
80
SWE-bench Verified
46
79
LiveCodeBench
38
85
SWE-bench Pro
43

Multimodal & Grounded

GPT-5.2

GPT-5.2

95

Mercury 2

68.3

95
MMMU-Pro
66
95
OfficeQA Pro
71

Reasoning

GPT-5.2

GPT-5.2

93.2

Mercury 2

80.1

95
SimpleQA
82
93
MuSR
82
96
BBH
87
91
LongBench v2
77
93
MRCRv2
76

Knowledge

GPT-5.2

GPT-5.2

79.5

Mercury 2

57.2

99
MMLU
78
97
GPQA
78
95
SuperGPQA
76
93
OpenBookQA
74
88
MMLU-Pro
72
42
HLE
9
91
FrontierScience
69

Instruction Following

GPT-5.2

GPT-5.2

94

Mercury 2

84

94
IFEval
84

Multilingual

GPT-5.2

GPT-5.2

92.4

Mercury 2

79.7

95
MGSM
81
91
MMLU-ProX
79

Mathematics

GPT-5.2

GPT-5.2

97.2

Mercury 2

80.9

99
AIME 2023
81
99
AIME 2024
83
98
AIME 2025
82
95
HMMT Feb 2023
77
97
HMMT Feb 2024
79
96
HMMT Feb 2025
78
96
BRUMO 2025
80
98
MATH-500
82

Frequently Asked Questions

Which is better, GPT-5.2 or Mercury 2?

GPT-5.2 is ahead overall, 88 to 65. The biggest single separator in this matchup is SWE-bench Pro, where the scores are 85 and 43.

Which is better for knowledge tasks, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for knowledge tasks in this comparison, averaging 79.5 versus 57.2. Inside this category, HLE is the benchmark that creates the most daylight between them.

Which is better for coding, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for coding in this comparison, averaging 81.8 versus 41.1. Inside this category, SWE-bench Pro is the benchmark that creates the most daylight between them.

Which is better for math, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for math in this comparison, averaging 97.2 versus 80.9. Inside this category, AIME 2023 is the benchmark that creates the most daylight between them.

Which is better for reasoning, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for reasoning in this comparison, averaging 93.2 versus 80.1. Inside this category, MRCRv2 is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for agentic tasks in this comparison, averaging 85.4 versus 63.7. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.

Which is better for multimodal and grounded tasks, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for multimodal and grounded tasks in this comparison, averaging 95 versus 68.3. Inside this category, MMMU-Pro is the benchmark that creates the most daylight between them.

Which is better for instruction following, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for instruction following in this comparison, averaging 94 versus 84. Inside this category, IFEval is the benchmark that creates the most daylight between them.

Which is better for multilingual tasks, GPT-5.2 or Mercury 2?

GPT-5.2 has the edge for multilingual tasks in this comparison, averaging 92.4 versus 79.7. Inside this category, MGSM is the benchmark that creates the most daylight between them.

Last updated: March 12, 2026

Weekly LLM Benchmark Digest

Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.

Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.