Atlas

Models

← All models

Gemini 2.5 Pro Preview 06-05

Google DeepMind · 2025-06-05

Gemini 2.5 Pro Preview 06-05 is Google's June 5, 2025 upgraded preview of Gemini 2.5 Pro. It added configurable thinking budgets for control over cost and latency; Google promoted this same version unchanged to the stable `gemini-2.5-pro` endpoint on June 17, 2025.

Benchmark scores

BenchmarkScore
Aider Polyglot83.1
BuseyBench SVG3.9
Capability146.9
Creative Writing v3 Elo1374
Creative Writing v3 Rubric Score16.2
DeepResearchBench0.4
Elimination Game TrueSkill Mu3.7
EnigmaEval5.6
Epoch Capabilities Index (ECI)146
EQ-Bench 31545
EQ-Bench 3 (Rubric Score)72.2
EQ-Bench 3: Analytic8.5
EQ-Bench 3: Assertive5.0
EQ-Bench 3: Compliant6.8
EQ-Bench 3: Empathy8.1
EQ-Bench 3: Humanlike6.4
EQ-Bench 3: Insight8.0
EQ-Bench 3: Moralising3.4
EQ-Bench 3: Pragmatic7.5
EQ-Bench 3: Safety5.5
EQ-Bench 3: Social IQ7.1
EQ-Bench 3: Warm7.8
Fiction.liveBench91.7
FORTRESS Average Risk Score (ARS)61.7
GPQA Diamond84.8
GSO Opt@13.9
HardGeoBench <=100mi48.8
HardGeoBench <=10mi37.5
HardGeoBench Avg Distance1119
HardGeoBench Country Accuracy56.3
Loading Atlas data…