Google DeepMindproprietary
Released: 2025-03-25Gemini 2.5 Pro
Google DeepMind frontier multimodal model supporting 2M context window with integrated thinking and top-tier complex coding performance.
Intelligence Index
91.2/ 100
Calibrated multi-domain compositeThroughput Speed
85tok/sec
Output streaming throughputTime To First Token
320ms
Initial chunk server latencyLiveBench Contam-Free
81.0%
September 2026 suite scoreLiveBench Multi-Domain Evaluation Breakdown
Monthly refreshed contamination-free test setsLogical Reasoning86.5%
Coding & Repo Repair82.0%
Mathematics (AIME)86.0%
Data Analysis & Tables78.5%
Language & Comprehension76.2%
Instruction Following77.0%
Sourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| ARC-AGI-3 | Active | 57.2% | 93 | 2025-04-29 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 68.6% | 75 | 2025-04-22 | independent | Source ↗ |
| FrontierMath | Active | 25.1% | 98 | 2025-04-19 | independent | Source ↗ |
| PutnamBench | Active | 46.5% | 90 | 2025-04-18 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 58.8% | 20 | 2025-04-16 | independent | Source ↗ |
| Omni-MATH | Deprecated | 66.2% | 10 | 2025-04-16 | independent | Source ↗ |
| Soohak | Active | 68% | 98 | 2025-04-14 | independent | Source ↗ |
| FrontierCode | Active | 50.2% | 98 | 2025-04-14 | independent | Source ↗ |
| Humanity's Last Exam | Active | 38.7% | 93 | 2025-04-12 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 65.5% | 75 | 2025-04-12 | independent | Source ↗ |
| AIME | Nearing Saturation | 85% | 75 | 2025-04-12 | independent | Source ↗ |
| OlympiadBench | Active | 73.4% | 80 | 2025-04-11 | independent | Source ↗ |
| MathArena | Active | 71.6% | 99 | 2025-04-10 | independent | Source ↗ |
| FireBench | Active | 64.7% | 95 | 2025-04-10 | independent | Source ↗ |
| Terminal-Bench | Active | 60.6% | 80 | 2025-04-10 | independent | Source ↗ |
| OpenAI-MRCR | Active | 76.6% | 80 | 2025-04-09 | independent | Source ↗ |
| SimpleQA | Active | 47.4% | 90 | 2025-04-09 | independent | Source ↗ |
| CodeContests | Deprecated | 56.2% | 10 | 2025-04-09 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 63.1% | 80 | 2025-04-08 | independent | Source ↗ |
| SWE-bench Pro | Active | 53.6% | 90 | 2025-04-08 | independent | Source ↗ |
| LongBench-Pro | Active | 74.8% | 80 | 2025-04-08 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 73% | 78 | 2025-04-07 | independent | Source ↗ |
| MultiPL-E | Deprecated | 79.6% | 10 | 2025-04-07 | independent | Source ↗ |
| LiveCodeBench | Saturated | 70.9% | 39 | 2025-04-06 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 52.4% | 80 | 2025-04-06 | independent | Source ↗ |
| MATH | Saturated | 87.7% | 25 | 2025-04-06 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 77.5% | 65 | 2025-04-06 | independent | Source ↗ |
| Mercury | Deprecated | 76.1% | 10 | 2025-04-05 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 73% | 65 | 2025-04-05 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 83.9% | 75 | 2025-04-05 | independent | Source ↗ |
| RULER | Nearing Saturation | 83.9% | 84 | 2025-04-04 | independent | Source ↗ |
| TabMWP | Saturated | 82.3% | 25 | 2025-04-04 | independent | Source ↗ |
| CyberSecEval | Deprecated | 77.9% | 10 | 2025-04-04 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 67.5% | 25 | 2025-04-03 | independent | Source ↗ |
| MMLU-Redux | Saturated | 89.4% | 25 | 2025-04-03 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 80.3% | 78 | 2025-04-03 | independent | Source ↗ |
| LongBench | Nearing Saturation | 78.4% | 65 | 2025-04-02 | independent | Source ↗ |
| SWE-bench | Deprecated | 58.8% | 10 | 2025-04-02 | independent | Source ↗ |
| MGSM | Nearing Saturation | 85% | 65 | 2025-04-02 | independent | Source ↗ |
| AGIEval | Saturated | 75.7% | 25 | 2025-04-01 | independent | Source ↗ |
| QASPER | Nearing Saturation | 75.7% | 65 | 2025-04-01 | independent | Source ↗ |
| TruthfulQA | Saturated | 80.3% | 25 | 2025-04-01 | independent | Source ↗ |
| GSM-Hard | Deprecated | 85.9% | 10 | 2025-03-31 | independent | Source ↗ |
| SuperGLUE | Saturated | 89.8% | 25 | 2025-03-31 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 81.3% | 23 | 2025-03-30 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 7.9% | 75 | 2025-03-30 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 86.2% | 25 | 2025-03-30 | independent | Source ↗ |
| WinoGrande | Saturated | 91.5% | 28 | 2025-03-29 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 88.2% | 65 | 2025-03-29 | independent | Source ↗ |
| IFEval | Deprecated | 75.5% | 25 | 2025-03-29 | independent | Source ↗ |
| GPQA | Nearing Saturation | 80.3% | 78 | 2025-03-29 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 98.2% | 44 | 2025-03-28 | independent | Source ↗ |
| HumanEval+ | Saturated | 93.4% | 25 | 2025-03-28 | independent | Source ↗ |
| CommonsenseQA | Saturated | 92.4% | 28 | 2025-03-28 | independent | Source ↗ |
| MBPP | Saturated | 91.7% | 25 | 2025-03-27 | independent | Source ↗ |
| GSM8K | Saturated | 96.7% | 25 | 2025-03-27 | independent | Source ↗ |
| GLUE | Saturated | 95% | 28 | 2025-03-27 | independent | Source ↗ |
| MMLU | Saturated | 93% | 25 | 2025-03-26 | independent | Source ↗ |
| HumanEval | Saturated | 98.6% | 25 | 2025-03-26 | independent | Source ↗ |
| ARC-AGI | Saturated | 82.7% | 25 | 2025-03-26 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 81% | 94 | 2025-02-01 | independent | Source ↗ |
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gemini-2-5-pro",
messages=[{"role": "user", "content": "Evaluate multi-step logic problem"}],
temperature=0.2
)
print(response.choices[0].message.content)Evaluation & Sourcing Notes
Scores listed for Gemini 2.5 Pro represent verified evaluations extracted from official research papers, independent evaluation suites (HELM, LMSYS, OpenCompass, LiveBench), and verified audit reports.
All benchmarks marked as saturated or deprecated reflect historical performance where the benchmark no longer provides active discriminative power.