Which is better, Gemini 3.5 Flash or GPT-5.2?

Across 45 shared benchmarks tracked on Vector Wire, Gemini 3.5 Flash leads on 31 and GPT-5.2 leads on 14 — Gemini 3.5 Flash leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Gemini 3.5 Flash vs GPT-5.2

GooglevsOpenAI45 shared benchmarks31–14 head-to-head

Benchmark	Gemini 3.5 Flash	GPT-5.2
AA Agentic Index	70.4	60.2
AA Intelligence	52	43.3
AA-LCR	81	79.3
AA-Omniscience	21.2	0.3
aime_2026	95	98.3
apexAgents	47.1	23
arc_agi_1	92.5	86.2
ARC-AGI-2	72.1	52.9
arena_elo	1477	1437
arena_text_factuality	1466	1442
Artificial Analysis Coding Index	70.1	48.7
charxiv_rq	84.2	82.1
coding_arena_elo	1490	1418
critpt	13.1	11.6
frontiermath_tier_4	14.6	18.8
gdpval	57.8	48.3
GDPval-AA (Elo)	1656	1462
GPQA Diamond	92.2	92.4
HLE	42.7	45.5
hmmt_feb_2026	95.5	97
IFBench	76.3	75.4
livebench	75	74.8
livebench_agentic_coding	49	50.3
livebench_coding	78.2	76.1
livebench_data_analysis	64.9	78.2
livebench_instruction_following	75.6	61.8
livebench_language	84.6	79.8
livebench_math	88.2	93.2
livebench_reasoning	82	83.2
matharena_visual_math_overall	89.9	86.5
MCP Atlas	83.6	68
MMMU-Pro	84.3	79.5
mrcr_v2_8needle_128k_average	77.3	83.8
OmniScience Accuracy	51.4	44.3
OmniScience Non-Hallucination	38.2	38.4
scicode	53.1	52.1
simplebench	76.7	45.8
simpleqa_verified	68.4	38.9
SWE-bench Pro	55.1	55.6
TauBench V3 - Banking	32.2	11.1
Terminal-Bench 2.0	76.2	54
Terminal-Bench Hard	46.2	62.2
toolathlon	56.5	46.3
τ²-Bench Telecom (AA run)	95.6	98.7
τ³-Bench	25.4	12

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.