Compare · ModelsLive · 2 picked · head to head

Gemini 3.5 Flash vs GPT-5.2

Side by side · benchmarks, pricing, and signals you can act on.

Winner summary

Gemini 3.5 Flash wins 10 of 16 shared benchmarks. Leads in agentic · reasoning · arena.

Category leads
agentic·Gemini 3.5 Flashreasoning·Gemini 3.5 Flasharena·Gemini 3.5 Flashknowledge·Gemini 3.5 Flashmath·GPT-5.2coding·Gemini 3.5 Flash
Hype vs Reality
Gemini 3.5 Flash
#44 by perf·no signal
QUIET
GPT-5.2
#93 by perf·no signal
QUIET
Best value
1.8x better value than GPT-5.2
Gemini 3.5 Flash
11.9 pts/$
$5.25/M
GPT-5.2
6.8 pts/$
$7.88/M
Vendor risk
Google DeepMind logo
Google DeepMind
$4.00T·Tier 1
Low risk
OpenAI logo
OpenAI
$840.0B·Tier 1
Medium risk
Head to head
Gemini 3.5 FlashGPT-5.2
APEX-Agents
Gemini 3.5 Flash leads by +15.3
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
Gemini 3.5 Flash
49.6
GPT-5.2
34.3
ARC-AGI
Gemini 3.5 Flash leads by +6.3
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
Gemini 3.5 Flash
92.5
GPT-5.2
86.2
ARC-AGI-2
Gemini 3.5 Flash leads by +19.2
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
Gemini 3.5 Flash
72.1
GPT-5.2
52.9
Chatbot Arena Elo · Coding
Gemini 3.5 Flash leads by +100.6
Gemini 3.5 Flash
1505.5
GPT-5.2
1405.0
Chatbot Arena Elo · Overall
Gemini 3.5 Flash leads by +41.6
Gemini 3.5 Flash
1476.3
GPT-5.2
1434.6
Chess Puzzles
Gemini 3.5 Flash leads by +1.0
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
Gemini 3.5 Flash
50.0
GPT-5.2
49.0
FrontierMath-2025-02-28-Private
GPT-5.2 leads by +1.7
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
Gemini 3.5 Flash
39.0
GPT-5.2
40.7
FrontierMath-Tier-4-2025-07-01-Private
GPT-5.2 leads by +4.2
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
Gemini 3.5 Flash
14.6
GPT-5.2
18.8
FrontierMath-Tier-4-v2-Private
GPT-5.2 leads by +4.9
Gemini 3.5 Flash
26.8
GPT-5.2
31.7
FrontierMath-Tiers-1-3-v2-Private
GPT-5.2 leads by +4.6
Gemini 3.5 Flash
62.8
GPT-5.2
67.4
GPQA diamond
Gemini 3.5 Flash leads by +1.9
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
Gemini 3.5 Flash
90.4
GPT-5.2
88.5
OTIS Mock AIME 2024-2025
GPT-5.2 leads by +0.6
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
Gemini 3.5 Flash
95.5
GPT-5.2
96.1
SimpleBench
Gemini 3.5 Flash leads by +37.1
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
Gemini 3.5 Flash
72.0
GPT-5.2
35.0
SimpleQA Verified
Gemini 3.5 Flash leads by +29.5
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
Gemini 3.5 Flash
68.4
GPT-5.2
38.9
SWE-Bench verified
Gemini 3.5 Flash leads by +5.6
SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. Claude Mythos Preview leads at 93.9%, crossing 90% for the first time in 2026. Opus 4.6 scores 80.8%. The benchmark remains the most-cited evaluation for code-generation capability.
Gemini 3.5 Flash
79.3
GPT-5.2
73.8
WeirdML
GPT-5.2 leads by +9.6
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
Gemini 3.5 Flash
62.6
GPT-5.2
72.2
Full benchmark table
BenchmarkGemini 3.5 FlashGPT-5.2
APEX-Agents
APEX-Agents · evaluates AI agents on complex, multi-step tasks requiring planning, tool use, and autonomous decision-making in realistic environments.
49.634.3
ARC-AGI
ARC-AGI · the original Abstraction and Reasoning Corpus, testing whether AI can solve novel visual pattern recognition tasks without memorization.
92.586.2
ARC-AGI-2
ARC-AGI-2 · the second iteration of the Abstraction and Reasoning Corpus, testing novel pattern recognition and abstract reasoning without prior training data.
72.152.9
Chatbot Arena Elo · Coding
1505.51405.0
Chatbot Arena Elo · Overall
1476.31434.6
Chess Puzzles
Chess Puzzles · tests strategic and tactical reasoning by having models solve chess puzzle positions, evaluating lookahead and pattern recognition abilities.
50.049.0
FrontierMath-2025-02-28-Private
FrontierMath (Feb 2025) · original research-level math problems created by mathematicians, testing capabilities at the boundary of current AI mathematical reasoning.
39.040.7
FrontierMath-Tier-4-2025-07-01-Private
FrontierMath Tier 4 (Jul 2025) · the most challenging tier of frontier mathematics, containing problems that push the absolute limits of AI mathematical reasoning.
14.618.8
FrontierMath-Tier-4-v2-Private
26.831.7
FrontierMath-Tiers-1-3-v2-Private
62.867.4
GPQA diamond
Graduate-Level Google-Proof QA (Diamond set) · expert-crafted questions in physics, biology, and chemistry that are difficult even for domain PhDs.
90.488.5
OTIS Mock AIME 2024-2025
OTIS Mock AIME 2024-2025 · simulated American Invitational Mathematics Examination problems testing advanced problem-solving skills.
95.596.1
SimpleBench
SimpleBench · tests fundamental reasoning capabilities with straightforward problems designed to expose gaps in basic logical and spatial thinking.
72.035.0
SimpleQA Verified
SimpleQA Verified · short factual questions with verified answers, measuring factual accuracy and the tendency to hallucinate or provide incorrect information.
68.438.9
SWE-Bench verified
SWE-bench Verified · 500 human-validated tasks from 12 real Python repositories (Django, Flask, scikit-learn, sympy, and others). Each task requires the model to produce a git patch that resolves a real GitHub issue and passes the test suite. The verified subset eliminates ambiguous tasks from the original SWE-bench. Claude Mythos Preview leads at 93.9%, crossing 90% for the first time in 2026. Opus 4.6 scores 80.8%. The benchmark remains the most-cited evaluation for code-generation capability.
79.373.8
WeirdML
WeirdML · tests models on unusual and adversarial machine learning tasks that require creative problem-solving beyond standard patterns.
62.672.2
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
Google DeepMind logoGemini 3.5 Flash$1.50$9.001.0M tokens (~524 books)$33.75
OpenAI logoGPT-5.2$1.75$14.00400K tokens (~200 books)$48.13