Which is better, Claude Sonnet 4.5 or Gemini 3 Flash Preview?

Across 46 shared benchmarks tracked on Vector Wire, Claude Sonnet 4.5 leads on 13 and Gemini 3 Flash Preview leads on 32 — Gemini 3 Flash Preview leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Sonnet 4.5 vs Gemini 3 Flash Preview

AnthropicvsGoogle46 shared benchmarks13–32 head-to-head

Benchmark	Claude Sonnet 4.5	Gemini 3 Flash Preview
AA Agentic Index	50.6	49.7
AA Intelligence	37.4	38.7
AA-LCR	68.3	73
AA-Omniscience	-0.1	10.1
AIME 2025	88	99.7
arc_agi_1	63.7	84.7
ARC-AGI-2	13.6	33.6
arena_elo	1456	1473
arena_text_factuality	1448	1452
Artificial Analysis Coding Index	52.1	42.6
charxiv_rq	67.2	80.3
coding_arena_elo	1392	1387
critpt	1.1	8.6
frontiermath_tier_4	4.2	4.2
gdpval	40.4	35.2
GPQA Diamond	83.4	90.4
HLE	19.8	43.5
hmmt_feb_2025	79.2	97.5
hmmt_nov_2025	81.7	93.3
IFBench	57.3	78
matharena_visual_math_overall	75.8	85.8
MCP Atlas	59.5	62
mmmlu	89.1	91.8
MMMU-Pro	68.7	81.2
OmniDocBench 1.5	85.8	12.1
OmniScience Accuracy	32.9	53.4
OmniScience Non-Hallucination	50.9	7.6
OSWorld-Verified	61.4	65.1
scicode	45	50.6
simplebench	54.3	61.1
simpleqa_verified	23.6	67.4
swe_bench_bash	71.4	75.8
SWE-bench Multilingual	68	72.7
SWE-bench Pro	43.6	49.6
SWE-bench Verified	82	78
TauBench V3 - Banking	24.5	20.8
Terminal-Bench 2.0	50	47.6
Terminal-Bench 2.1	55.8	58
Terminal-Bench Hard	35.6	38.6
toolathlon	41	49.4
vectara_answer_rate	95.6	99.8
vectara_avg_summary_length	127.8	90.2
vectara_factual_consistency	88	86.5
vectara_hallucination_rate ↓	12	13.5
τ²-Bench Telecom (AA run)	78.1	80.4
τ³-Bench	19	17.5

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.