Google DeepMindproprietary
Released: 2026-09-02Gemini 3.8 Flash
Google high-throughput intelligence workhorse model delivering 185 tok/s with 1M context multimodal reasoning.
Intelligence Index
93.8/ 100
Calibrated multi-domain compositeThroughput Speed
185tok/sec
Output streaming throughputTime To First Token
230ms
Initial chunk server latencyLiveBench Contam-Free
80.2%
September 2026 suite scoreSourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| MultiPL-E | Deprecated | 74.5% | 10 | 2026-09-12 | independent | Source ↗ |
| Humanity's Last Exam | Active | 41.1% | 93 | 2026-09-12 | independent | Source ↗ |
| SWE-bench Pro | Active | 50.2% | 90 | 2026-09-12 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 79.7% | 65 | 2026-09-12 | independent | Source ↗ |
| LongBench-Pro | Active | 76.9% | 80 | 2026-09-12 | independent | Source ↗ |
| MATH | Saturated | 92.1% | 25 | 2026-09-12 | independent | Source ↗ |
| MathArena | Active | 75.2% | 99 | 2026-09-12 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 58.7% | 80 | 2026-09-12 | independent | Source ↗ |
| Mercury | Deprecated | 71.3% | 10 | 2026-09-12 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 86.3% | 75 | 2026-09-12 | independent | Source ↗ |
| LiveCodeBench | Saturated | 66.4% | 39 | 2026-09-12 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 73.4% | 75 | 2026-09-12 | independent | Source ↗ |
| OlympiadBench | Active | 77.1% | 80 | 2026-09-12 | independent | Source ↗ |
| Omni-MATH | Deprecated | 69.6% | 10 | 2026-09-12 | independent | Source ↗ |
| OpenAI-MRCR | Active | 78.8% | 80 | 2026-09-12 | independent | Source ↗ |
| PutnamBench | Active | 48.9% | 90 | 2026-09-12 | independent | Source ↗ |
| RULER | Nearing Saturation | 86.3% | 84 | 2026-09-12 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 75% | 65 | 2026-09-12 | independent | Source ↗ |
| SimpleQA | Active | 48.8% | 90 | 2026-09-12 | independent | Source ↗ |
| Soohak | Active | 71.4% | 98 | 2026-09-12 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 55.1% | 20 | 2026-09-12 | independent | Source ↗ |
| AIME | Nearing Saturation | 89.3% | 75 | 2026-09-12 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 75% | 78 | 2026-09-12 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 72.9% | 75 | 2026-09-12 | independent | Source ↗ |
| ARC-AGI-3 | Active | 60.8% | 93 | 2026-09-12 | independent | Source ↗ |
| CodeContests | Deprecated | 52.7% | 10 | 2026-09-12 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 70.8% | 80 | 2026-09-12 | independent | Source ↗ |
| CyberSecEval | Deprecated | 72.9% | 10 | 2026-09-12 | independent | Source ↗ |
| FireBench | Active | 72.5% | 95 | 2026-09-12 | independent | Source ↗ |
| FrontierCode | Active | 47% | 98 | 2026-09-12 | independent | Source ↗ |
| FrontierMath | Active | 26.3% | 98 | 2026-09-12 | independent | Source ↗ |
| TabMWP | Saturated | 86.5% | 25 | 2026-09-12 | independent | Source ↗ |
| Terminal-Bench | Active | 56.7% | 80 | 2026-09-12 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 63.2% | 25 | 2026-09-11 | independent | Source ↗ |
| MMLU-Redux | Saturated | 91.9% | 25 | 2026-09-11 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 82.5% | 78 | 2026-09-11 | independent | Source ↗ |
| MGSM | Nearing Saturation | 89.3% | 65 | 2026-09-10 | independent | Source ↗ |
| SWE-bench | Deprecated | 55.1% | 10 | 2026-09-10 | independent | Source ↗ |
| LongBench | Nearing Saturation | 80.7% | 65 | 2026-09-10 | independent | Source ↗ |
| AGIEval | Saturated | 80.4% | 25 | 2026-09-09 | independent | Source ↗ |
| QASPER | Nearing Saturation | 77.9% | 65 | 2026-09-09 | independent | Source ↗ |
| TruthfulQA | Saturated | 82.5% | 25 | 2026-09-09 | independent | Source ↗ |
| GSM-Hard | Deprecated | 90.2% | 10 | 2026-09-08 | independent | Source ↗ |
| SuperGLUE | Saturated | 95.4% | 25 | 2026-09-08 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 91.6% | 25 | 2026-09-07 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 76.1% | 23 | 2026-09-07 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 8.9% | 75 | 2026-09-07 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 82.6% | 65 | 2026-09-06 | independent | Source ↗ |
| WinoGrande | Saturated | 97.2% | 28 | 2026-09-06 | independent | Source ↗ |
| IFEval | Deprecated | 84.6% | 25 | 2026-09-06 | independent | Source ↗ |
| GPQA | Nearing Saturation | 82.5% | 78 | 2026-09-06 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 98.8% | 44 | 2026-09-05 | independent | Source ↗ |
| HumanEval+ | Saturated | 87.5% | 25 | 2026-09-05 | independent | Source ↗ |
| CommonsenseQA | Saturated | 98.2% | 28 | 2026-09-05 | independent | Source ↗ |
| MBPP | Saturated | 85.9% | 25 | 2026-09-04 | independent | Source ↗ |
| GSM8K | Saturated | 100% | 25 | 2026-09-04 | independent | Source ↗ |
| GLUE | Saturated | 100% | 28 | 2026-09-04 | independent | Source ↗ |
| MMLU | Saturated | 95.7% | 25 | 2026-09-03 | independent | Source ↗ |
| HumanEval | Saturated | 92.3% | 25 | 2026-09-03 | independent | Source ↗ |
| ARC-AGI | Saturated | 87.9% | 25 | 2026-09-03 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 80.2% | 94 | 2025-02-01 | independent | Source ↗ |
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gemini-3-8-flash",
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 3.8 Flash 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.