Which is better, Gemini 3 Flash Preview or GPT-5.5?

Across 47 shared benchmarks tracked on Vector Wire, Gemini 3 Flash Preview leads on 3 and GPT-5.5 leads on 44 — GPT-5.5 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 Flash Preview vs GPT-5.5

GooglevsOpenAI47 shared benchmarks3–44 head-to-head

Benchmark	Gemini 3 Flash Preview	GPT-5.5
AA Agentic Index	49.7	47.4
AA Intelligence	38.7	56.3
AA-LCR	73	79
AA-Omniscience	10.1	20.5
aime_2026	96.7	100
apexAgents	27.7	37.7
arc_agi_1	84.7	95
ARC-AGI-2	33.6	85
arena_elo	1473	1482
arena_text_factuality	1452	1482
arena_vision	1260	1283
Artificial Analysis Coding Index	42.6	74.9
coding_arena_elo	1387	1457
critpt	8.6	27.1
Finance Agent v2	42.6	51.8
frontiermath_tier_4	4.2	35.4
gdpval	35.2	49.5
GDPval-AA (Elo)	1204	1769
GPQA Diamond	90.4	93.6
HLE	43.5	52.2
hmmt_feb_2026	89.4	98.5
hmmt_nov_2025	93.3	96.5
IFBench	78	75.9
Legal Agent Benchmark	0	2.1
livebench	72.4	80.7
matharena_visual_math_overall	85.8	94.9
MCP Atlas	62	82.8
MMMU-Pro	81.2	83.2
OmniScience Accuracy	53.4	58
OmniScience Non-Hallucination	7.6	12
OSWorld-Verified	65.1	78.7
scicode	50.6	56.1
simplebench	61.1	69
simpleqa_verified	67.4	63.1
SWE-bench Pro	49.6	58.6
SWE-bench Verified	78	80.6
TauBench V3 - Banking	20.8	39
Terminal-Bench 2.0	47.6	82.7
Terminal-Bench 2.1	58	84.3
Terminal-Bench Hard	38.6	60.6
toolathlon	49.4	55.6
vectara_answer_rate	99.8	100
vectara_avg_summary_length	90.2	129.6
vectara_factual_consistency	86.5	90.7
vectara_hallucination_rate ↓	13.5	9.3
τ²-Bench Telecom (AA run)	80.4	93.9
τ³-Bench	17.5	31.3

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.