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

Across 46 shared benchmarks tracked on Vector Wire, Claude Opus 4.5 leads on 25 and Gemini 3 Flash Preview leads on 19 — Claude Opus 4.5 leads on more. Strengths differ by axis; the table below shows every head-to-head number with provenance.

SEEDRefresh Models —Deals —Regulatory —Sources —

Claude Opus 4.5 vs Gemini 3 Flash Preview

AnthropicvsGoogle46 shared benchmarks25–19 head-to-head

Benchmark	Claude Opus 4.5	Gemini 3 Flash Preview
AA Agentic Index	59.6	49.7
AA Intelligence	41.9	38.7
AA-LCR	76	73
AA-Omniscience	14	10.1
AIME 2025	92.8	99.7
aime_2026	95.1	96.7
arc_agi_1	82	84.7
ARC-AGI-2	37.6	33.6
arena_elo	1473	1473
arena_text_factuality	1458	1452
Artificial Analysis Coding Index	47.8	42.6
charxiv_rq	68.5	80.3
coding_arena_elo	1495	1387
critpt	4.6	8.6
frontiermath_tier_4	4.2	4.2
gdpval	47.3	35.2
GPQA Diamond	87	90.4
HLE	30.8	43.5
hmmt_feb_2025	92.9	97.5
hmmt_feb_2026	85.3	89.4
hmmt_nov_2025	93.3	93.3
IFBench	58	78
livebench	76	72.4
MCP Atlas	69.8	62
mmmlu	90.8	91.8
MMMU-Pro	74	81.2
OmniDocBench 1.5	87.7	12.1
OmniScience Accuracy	46.6	53.4
OmniScience Non-Hallucination	39	7.6
OSWorld-Verified	66.3	65.1
scicode	50	50.6
simplebench	62	61.1
simpleqa_verified	41.8	67.4
swe_bench_bash	76.8	75.8
SWE-bench Multilingual	77.5	72.7
SWE-bench Pro	57.1	49.6
SWE-bench Verified	81.5	78
Terminal-Bench 2.0	59.3	47.6
Terminal-Bench Hard	47	38.6
toolathlon	43.5	49.4
vectara_answer_rate	98.7	99.8
vectara_avg_summary_length	114.5	90.2
vectara_factual_consistency	89.1	86.5
vectara_hallucination_rate ↓	10.9	13.5
VideoMMMU	84.4	86.9
τ²-Bench Telecom (AA run)	89.5	80.4

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.