Google DeepMindproprietary
Released: 2026-05-18

Gemini 3 Pro

Google DeepMind multimodal frontier model with native video comprehension and 2M token context window.

Intelligence Index
92.5/ 100
Calibrated multi-domain composite
Throughput Speed
80tok/sec
Output streaming throughput
Time To First Token
360ms
Initial chunk server latency
LiveBench Contam-Free
79.8%
September 2026 suite score

LiveBench Multi-Domain Evaluation Breakdown

Monthly refreshed contamination-free test sets
Logical Reasoning92.0%
Coding & Repo Repair81.5%
Mathematics (AIME)92.0%
Data Analysis & Tables88.0%
Language & Comprehension89.0%
Instruction Following93.5%

Sourced Benchmark Evaluations (4)

Verified performance across authoritative benchmarks with provenance tracking.
BenchmarkStatusScoreTrustDateSource TypeProvenance
ARC-AGI-2Nearing Saturation71.8%
75
2026-06-15independentSource ↗
SWE-bench VerifiedSaturated61.9%
25
2026-05-27independentSource ↗
IFEvalDeprecated93.5%
25
2026-05-22independentSource ↗
Humanity's Last ExamActive38.3%
90
2024-06-01independentSource ↗
OpenAI Compatible API ExecutionAPI Ready
from openai import OpenAI

client = OpenAI()
response = client.chat.completions.create(
    model="gemini-3-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 3 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.