Gemini 2.5 Pro vs grok-4

Every finance benchmark where both have a score (Google and xAI). Data refreshed 2 Nov 2026.

Shared benchmarks
2
Gemini 2.5 Pro only
3
grok-4 only
3

Shared benchmarks, side by side

Each benchmark is drawn on its own scale. Bars are not comparable across rows.

  • Gemini 2.5 Pro
  • grok-4
Benchmark
Gemini 2.5 Pro
grok-4

10 benchmarks, 3 scored for both. Bold marks the better score where both have one.

Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation)Level I accuracy %, higher is better
95.7
Paper Table 5 (Reported by the authors, 9 Dec 2025)
94.8
Paper Table 5 (Reported by the authors, 9 Dec 2025)
Reasoning Models Ace the CFA Exams (CFA mock-exam evaluation)Level III constructed-response score %, higher is better
78.2
Paper Table 5 (Reported by the authors, 9 Dec 2025)
71.2
Paper Table 5 (Reported by the authors, 9 Dec 2025)
FinDeepResearchoverall accuracy score (0-100), higher is better
36.2
OpenFinArena leaderboard (Official board, 15 Oct 2025)
37.3
OpenFinArena leaderboard (Official board, 15 Oct 2025)
Market-Benchmean MAE, lower is better
Not scored
443.2
marketbench.ai leaderboard (Official board, date not stated)
Advanced Financial Reasoning at Scale (CFA Level III evaluation)overall % (mean of MCQ and essay), higher is better
75.9
Paper Table 1 (Reported by the authors, 22 Sept 2025)
Not scored
FAITHoverall accuracy %, higher is better
91.9
Paper Table 3 (Reported by the authors, 24 Oct 2025)
Not scored
FinSearchCompGlobal subset average accuracy %, higher is better
Not scored
68.9
Paper Table 5 (Reported by the authors, 16 Sept 2025)
FinSearchCompGreater China subset average accuracy %, higher is better
Not scored
51.9
Paper Table 5 (Reported by the authors, 16 Sept 2025)
K-FinHallubinary detection F1, higher is better
0.849
Paper Table 3 (Reported by the authors, 28 May 2026)
Not scored
LiveTradeBenchcumulative return %, higher is better
Not scored
4.3
Paper Table 2 (Reported by the authors, 5 Nov 2025)

Scores from different benchmarks use different scales and are not added up. Settings can differ between sources; see each leaderboard for details.