Google DeepMindproprietary
Released: 2026-06-01Google: Gemma 2 27B
Gemma 2 27B by Google is an open model built from the same research and technology used to create the [Gemini models](/models?q=gemini). Gemma models are well-suited for a variety of...
Intelligence Index
81.0/ 100
Calibrated multi-domain compositeThroughput Speed
130tok/sec
Output streaming throughputTime To First Token
280ms
Initial chunk server latencyLiveBench Contam-Free
71.3%
September 2026 suite scoreLiveBench Multi-Domain Evaluation Breakdown
Monthly refreshed contamination-free test setsLogical Reasoning74.5%
Coding & Repo Repair72.9%
Mathematics (AIME)72.1%
Data Analysis & Tables68.9%
Language & Comprehension70.5%
Instruction Following71.3%
Sourced Benchmark Evaluations (51)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| ARC-AGI-3 | Active | 51.1% | 93 | 2026-07-06 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 61.3% | 75 | 2026-06-29 | independent | Source ↗ |
| FrontierMath | Active | 21.1% | 98 | 2026-06-26 | independent | Source ↗ |
| PutnamBench | Active | 39.2% | 90 | 2026-06-25 | independent | Source ↗ |
| Omni-MATH | Deprecated | 55.7% | 10 | 2026-06-23 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 52.4% | 20 | 2026-06-23 | independent | Source ↗ |
| Soohak | Active | 57.2% | 98 | 2026-06-21 | independent | Source ↗ |
| FrontierCode | Active | 44.7% | 98 | 2026-06-21 | independent | Source ↗ |
| AIME | Nearing Saturation | 71.5% | 75 | 2026-06-19 | independent | Source ↗ |
| Humanity's Last Exam | Active | 34.6% | 93 | 2026-06-19 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 60.6% | 75 | 2026-06-19 | independent | Source ↗ |
| OlympiadBench | Active | 61.7% | 80 | 2026-06-18 | independent | Source ↗ |
| MathArena | Active | 60.2% | 99 | 2026-06-17 | independent | Source ↗ |
| FireBench | Active | 59.9% | 95 | 2026-06-17 | independent | Source ↗ |
| Terminal-Bench | Active | 53.9% | 80 | 2026-06-17 | independent | Source ↗ |
| CodeContests | Deprecated | 50.1% | 10 | 2026-06-16 | independent | Source ↗ |
| SimpleQA | Active | 42.1% | 90 | 2026-06-16 | independent | Source ↗ |
| SWE-bench Pro | Active | 47.7% | 90 | 2026-06-15 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 58.5% | 80 | 2026-06-15 | independent | Source ↗ |
| MultiPL-E | Deprecated | 70.8% | 10 | 2026-06-14 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 48.5% | 80 | 2026-06-13 | independent | Source ↗ |
| MATH | Saturated | 73.8% | 25 | 2026-06-13 | independent | Source ↗ |
| LiveCodeBench | Saturated | 63.1% | 39 | 2026-06-13 | independent | Source ↗ |
| Mercury | Deprecated | 67.8% | 10 | 2026-06-12 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 74.5% | 75 | 2026-06-12 | independent | Source ↗ |
| CyberSecEval | Deprecated | 69.3% | 10 | 2026-06-11 | independent | Source ↗ |
| TabMWP | Saturated | 69.3% | 25 | 2026-06-11 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 60.1% | 25 | 2026-06-10 | independent | Source ↗ |
| MMLU-Redux | Saturated | 79.4% | 25 | 2026-06-10 | independent | Source ↗ |
| SWE-bench | Deprecated | 52.4% | 10 | 2026-06-09 | independent | Source ↗ |
| MGSM | Nearing Saturation | 71.5% | 65 | 2026-06-09 | independent | Source ↗ |
| TruthfulQA | Saturated | 71.3% | 25 | 2026-06-08 | independent | Source ↗ |
| AGIEval | Saturated | 67.6% | 25 | 2026-06-08 | independent | Source ↗ |
| SuperGLUE | Saturated | 80.2% | 25 | 2026-06-07 | independent | Source ↗ |
| GSM-Hard | Deprecated | 72.3% | 10 | 2026-06-07 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 7.3% | 75 | 2026-06-06 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 72.4% | 23 | 2026-06-06 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 77% | 25 | 2026-06-06 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 78.5% | 65 | 2026-06-05 | independent | Source ↗ |
| IFEval | Deprecated | 69.9% | 25 | 2026-06-05 | independent | Source ↗ |
| WinoGrande | Saturated | 81.7% | 28 | 2026-06-05 | independent | Source ↗ |
| GPQA | Nearing Saturation | 71.3% | 78 | 2026-06-05 | independent | Source ↗ |
| HumanEval+ | Saturated | 83.2% | 25 | 2026-06-04 | independent | Source ↗ |
| CommonsenseQA | Saturated | 82.5% | 28 | 2026-06-04 | independent | Source ↗ |
| MBPP | Saturated | 81.6% | 25 | 2026-06-03 | independent | Source ↗ |
| GSM8K | Saturated | 81.3% | 25 | 2026-06-03 | independent | Source ↗ |
| GLUE | Saturated | 84.9% | 28 | 2026-06-03 | independent | Source ↗ |
| MMLU | Saturated | 82.6% | 25 | 2026-06-02 | independent | Source ↗ |
| ARC-AGI | Saturated | 73.9% | 25 | 2026-06-02 | independent | Source ↗ |
| HumanEval | Saturated | 87.8% | 25 | 2026-06-02 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 71.3% | 94 | 2025-02-01 | independent | Source ↗ |
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="google-gemma-2-27b",
messages=[{"role": "user", "content": "Evaluate multi-step logic problem"}],
temperature=0.2
)
print(response.choices[0].message.content)Evaluation & Sourcing Notes
Scores listed for Google: Gemma 2 27B 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.