Google DeepMindproprietary
Released: 2025-01-20Gemini 2.0 Flash Thinking
Google DeepMind reasoning model with native multimodal capabilities and 1M token context.
Intelligence Index
87.9/ 100
Calibrated multi-domain compositeThroughput Speed
95tok/sec
Output streaming throughputTime To First Token
500ms
Initial chunk server latencyLiveBench Contam-Free
74.9%
September 2026 suite scoreLiveBench Multi-Domain Evaluation Breakdown
Monthly refreshed contamination-free test setsLogical Reasoning87.0%
Coding & Repo Repair75.2%
Mathematics (AIME)88.5%
Data Analysis & Tables79.0%
Language & Comprehension83.0%
Instruction Following89.0%
Sourced Benchmark Evaluations (61)
Verified performance across authoritative benchmarks with provenance tracking.| Benchmark | Status | Score | Trust | Date | Source Type | Provenance |
|---|---|---|---|---|---|---|
| ARC-AGI-3 | Active | 56.6% | 93 | 2025-02-24 | independent | Source ↗ |
| ARC-AGI-2 | Nearing Saturation | 67.9% | 75 | 2025-02-17 | independent | Source ↗ |
| FrontierMath | Active | 24.8% | 98 | 2025-02-14 | independent | Source ↗ |
| PutnamBench | Active | 46% | 90 | 2025-02-13 | independent | Source ↗ |
| Omni-MATH | Deprecated | 65.5% | 10 | 2025-02-11 | independent | Source ↗ |
| SWE-Lancer | Deprecated | 51.1% | 20 | 2025-02-11 | independent | Source ↗ |
| Soohak | Active | 67.3% | 98 | 2025-02-09 | independent | Source ↗ |
| FrontierCode | Active | 43.6% | 98 | 2025-02-09 | independent | Source ↗ |
| Humanity's Last Exam | Active | 38.3% | 93 | 2025-02-07 | independent | Source ↗ |
| StructFlowBench | Nearing Saturation | 75.6% | 75 | 2025-02-07 | independent | Source ↗ |
| AIME | Nearing Saturation | 84.1% | 75 | 2025-02-07 | independent | Source ↗ |
| OlympiadBench | Active | 72.6% | 80 | 2025-02-06 | independent | Source ↗ |
| FireBench | Active | 74.8% | 95 | 2025-02-05 | independent | Source ↗ |
| Terminal-Bench | Active | 52.6% | 80 | 2025-02-05 | independent | Source ↗ |
| MathArena | Active | 70.8% | 99 | 2025-02-05 | independent | Source ↗ |
| OpenAI-MRCR | Active | 73.8% | 80 | 2025-02-04 | independent | Source ↗ |
| CodeContests | Deprecated | 48.9% | 10 | 2025-02-04 | independent | Source ↗ |
| SimpleQA | Active | 45.7% | 90 | 2025-02-04 | independent | Source ↗ |
| ComplexBench | Nearing Saturation | 73% | 80 | 2025-02-03 | independent | Source ↗ |
| SWE-bench Pro | Active | 46.6% | 90 | 2025-02-03 | independent | Source ↗ |
| LongBench-Pro | Active | 72.1% | 80 | 2025-02-03 | independent | Source ↗ |
| MultiPL-E | Deprecated | 69.2% | 10 | 2025-02-02 | independent | Source ↗ |
| ZeroSCROLLS | Nearing Saturation | 70.3% | 78 | 2025-02-02 | independent | Source ↗ |
| AgentIF | Nearing Saturation | 60.5% | 80 | 2025-02-01 | independent | Source ↗ |
| MATH | Saturated | 86.7% | 25 | 2025-02-01 | independent | Source ↗ |
| LiveCodeBench | Saturated | 61.7% | 39 | 2025-02-01 | independent | Source ↗ |
| InfiniteBench | Nearing Saturation | 74.7% | 65 | 2025-02-01 | independent | Source ↗ |
| LiveBench (Continuous) | Active | 74.9% | 94 | 2025-02-01 | independent | Source ↗ |
| MMLU-Pro | Nearing Saturation | 80.9% | 75 | 2025-01-31 | independent | Source ↗ |
| Mercury | Deprecated | 66.2% | 10 | 2025-01-31 | independent | Source ↗ |
| SCROLLS | Nearing Saturation | 70.3% | 65 | 2025-01-31 | independent | Source ↗ |
| RULER | Nearing Saturation | 80.9% | 84 | 2025-01-30 | independent | Source ↗ |
| TabMWP | Saturated | 81.4% | 25 | 2025-01-30 | independent | Source ↗ |
| CyberSecEval | Deprecated | 67.7% | 10 | 2025-01-30 | independent | Source ↗ |
| MMLU-Redux | Saturated | 86.1% | 25 | 2025-01-29 | independent | Source ↗ |
| QuALITY | Nearing Saturation | 77.4% | 78 | 2025-01-29 | independent | Source ↗ |
| SWE-bench Verified | Saturated | 58.7% | 25 | 2025-01-29 | independent | Source ↗ |
| MGSM | Nearing Saturation | 84.1% | 65 | 2025-01-28 | independent | Source ↗ |
| LongBench | Nearing Saturation | 75.6% | 65 | 2025-01-28 | independent | Source ↗ |
| SWE-bench | Deprecated | 51.1% | 10 | 2025-01-28 | independent | Source ↗ |
| AGIEval | Saturated | 74.8% | 25 | 2025-01-27 | independent | Source ↗ |
| QASPER | Nearing Saturation | 73% | 65 | 2025-01-27 | independent | Source ↗ |
| TruthfulQA | Saturated | 77.4% | 25 | 2025-01-27 | independent | Source ↗ |
| SuperGLUE | Saturated | 88.7% | 25 | 2025-01-26 | independent | Source ↗ |
| GSM-Hard | Deprecated | 85% | 10 | 2025-01-26 | independent | Source ↗ |
| MT-Bench | Nearing Saturation | 9.2% | 75 | 2025-01-25 | independent | Source ↗ |
| BIG-Bench Hard | Saturated | 85.3% | 25 | 2025-01-25 | independent | Source ↗ |
| CodeXGLUE | Deprecated | 70.7% | 23 | 2025-01-25 | independent | Source ↗ |
| MBPP+ | Nearing Saturation | 76.7% | 65 | 2025-01-24 | independent | Source ↗ |
| WinoGrande | Saturated | 90.5% | 28 | 2025-01-24 | independent | Source ↗ |
| IFEval | Deprecated | 87.2% | 25 | 2025-01-24 | independent | Source ↗ |
| GPQA | Nearing Saturation | 77.4% | 78 | 2025-01-24 | independent | Source ↗ |
| HumanEval+ | Saturated | 81.2% | 25 | 2025-01-23 | independent | Source ↗ |
| Needle-in-a-Haystack | Saturated | 97.6% | 44 | 2025-01-23 | independent | Source ↗ |
| CommonsenseQA | Saturated | 91.4% | 28 | 2025-01-23 | independent | Source ↗ |
| GSM8K | Saturated | 95.6% | 25 | 2025-01-22 | independent | Source ↗ |
| GLUE | Saturated | 94% | 28 | 2025-01-22 | independent | Source ↗ |
| MBPP | Saturated | 79.7% | 25 | 2025-01-22 | independent | Source ↗ |
| HumanEval | Saturated | 85.7% | 25 | 2025-01-21 | independent | Source ↗ |
| MMLU | Saturated | 89.7% | 25 | 2025-01-21 | independent | Source ↗ |
| ARC-AGI | Saturated | 81.8% | 25 | 2025-01-21 | independent | Source ↗ |
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gemini-2-0-flash-thinking",
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 2.0 Flash Thinking 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.