Which is better, Claude Opus 4.7 or Gemini 3 Pro?

Across 46 shared benchmarks tracked on Vector Wire, Claude Opus 4.7 leads on 34 and Gemini 3 Pro leads on 12 — Claude Opus 4.7 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 4.7 vs Gemini 3 Pro

AnthropicvsGoogle46 shared benchmarks34–12 head-to-head

Benchmark	Claude Opus 4.7	Gemini 3 Pro
AA Agentic Index	64.6	52
AA Intelligence	55	40.6
AA-LCR	75.3	73
AA-Omniscience	27.3	15.3
aime_2026	95.8	91.7
arc_agi_1	92	75
ARC-AGI-2	75.8	54
arena_elo	1494	1485
arena_text_factuality	1474	1481
arena_vision	1304	1290
Artificial Analysis Coding Index	73.6	46.5
browsecomp	79.8	59.2
charxiv_rq	82.1	81.4
coding_arena_elo	1557	1438
critpt	12	9.1
cybergym	73.8	39.9
frontiermath_tier_4	22.9	18.8
gdpval	58.6	34.2
GDPval-AA (Elo)	1753	1195
GPQA Diamond	94.2	91.9
HLE	54.7	45.8
HLE (with tools)	54.7	45.8
hmmt_feb_2026	93.9	86.4
IFBench	58.6	70.4
livebench	76.9	73.4
MCP Atlas	79.1	70.3
mmmlu	91.5	91.8
MMMU-Pro	78.8	81
mrcr_v2_8needle_128k_average	59.3	77
OmniScience Accuracy	48.9	55.8
OmniScience Non-Hallucination	57.7	10
scicode	54.5	56.1
screenspot_pro_no_tools	79.5	72.7
simplebench	61.7	76.4
simpleqa_verified	50.6	72.9
SWE-bench Multilingual	80.5	68.7
SWE-bench Pro	64.3	43.3
SWE-bench Verified	87.6	78
Terminal-Bench 2.0	69.4	54.2
Terminal-Bench Hard	54.5	56.9
toolathlon	59.3	36.4
vectara_answer_rate	98	99.4
vectara_avg_summary_length	149.1	101.9
vectara_factual_consistency	88	86.4
vectara_hallucination_rate ↓	12	13.6
τ²-Bench Telecom (AA run)	88.6	98

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.