Which is better, Claude Opus 4.7 or Gemini 3.5 Flash?

Across 53 shared benchmarks tracked on Vector Wire, Claude Opus 4.7 leads on 34 and Gemini 3.5 Flash leads on 19 — 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 3.5 Flash

AnthropicvsGoogle53 shared benchmarks34–19 head-to-head

Benchmark	Claude Opus 4.7	Gemini 3.5 Flash
AA Agentic Index	64.6	70.4
AA Intelligence	55	52
AA-LCR	75.3	81
AA-Omniscience	27.3	21.2
aime_2026	95.8	95
arc_agi_1	92	92.5
ARC-AGI-2	75.8	72.1
arena_elo	1494	1477
arena_text_factuality	1474	1466
Artificial Analysis Coding Index	73.6	70.1
Blueprint-Bench 2	24.5	33.6
charxiv_reasoning_with_tools	91	84.9
charxiv_rq	82.1	84.2
coding_arena_elo	1557	1490
critpt	12	13.1
Finance Agent v2	51.5	57.9
finance_agent	71.5	57.9
frontiermath_tier_4	22.9	14.6
gdpval	58.6	57.8
GDPval-AA (Elo)	1753	1656
GPQA Diamond	94.2	92.2
HiL-Bench	41.7	27.7
HLE	54.7	42.7
hmmt_feb_2026	93.9	95.5
Humanity’s Last Exam	46.9	40.2
IFBench	58.6	76.3
itbenchSre	46.7	40.3
Legal Agent Benchmark	7.1	0.8
livebench	76.9	75
livebench_agentic_coding	50.7	49
livebench_coding	82.1	78.2
livebench_data_analysis	78.3	64.9
livebench_instruction_following	66.7	75.6
livebench_language	77.9	84.6
livebench_math	92.8	88.2
livebench_reasoning	87.2	82
MCP Atlas	79.1	83.6
MMMU-Pro	78.8	84.3
mrcr_v2_8needle_128k_average	59.3	77.3
OmniScience Accuracy	48.9	51.4
OmniScience Non-Hallucination	57.7	38.2
OSWorld-Verified	82.8	78.4
scicode	54.5	53.1
simplebench	61.7	76.7
simpleqa_verified	50.6	68.4
SWE-bench Pro	64.3	55.1
TauBench V3 - Banking	34.6	32.2
Terminal-Bench 2.0	69.4	76.2
Terminal-Bench 2.1	83.1	78.7
Terminal-Bench Hard	54.5	46.2
toolathlon	59.3	56.5
τ²-Bench Telecom (AA run)	88.6	95.6
τ³-Bench	28.9	25.4

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.