Which is better, Gemini 3.1 Pro or Opus 5?

Across 44 shared benchmarks tracked on Vector Wire, Gemini 3.1 Pro leads on 9 and Opus 5 leads on 35 — Opus 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 3.1 Pro vs Opus 5

GooglevsAnthropic44 shared benchmarks9–35 head-to-head

Benchmark	Gemini 3.1 Pro	Opus 5
AA Agentic Index	23	59.2
AA Intelligence	47.7	63.1
AA-LCR	79	78.7
AA-Omniscience	32.9	37.1
arc_agi_1	98	97.5
arc_agi_3	0.4	30.2
ARC-AGI-2	77.1	90.4
arena_elo	1486	1487
arena_text_factuality	1471	1481
Artificial Analysis Coding Index	68.8	78
browsecomp	85.9	90.8
coding_arena_elo	1447	1662
critpt	17.7	29.1
DeepSWE 1.1	12	68.8
Finance Agent v2	43	58.6
gdpval	23.3	67.2
GDPval-AA (Elo)	1317	1827
GDPval-AA v2	962	1861
GPQA Diamond	94.3	93.7
HiL-Bench	35.3	57
HLE	51.4	64.7
HLE (with tools)	51.6	64.7
Legal Agent Benchmark	0	23
livebench_agentic_coding	44.1	65.2
livebench_coding	76.5	81.5
livebench_data_analysis	78.5	74.5
livebench_instruction_following	79.1	63.8
livebench_language	85.4	88.7
livebench_math	91	95.7
livebench_reasoning	84	91.2
MCP Atlas	78.2	85.8
MMMU-Pro	83	84.7
officeqa_pro	42.9	66.9
OmniScience Accuracy	55.3	60.9
OmniScience Non-Hallucination	50.1	40.5
OSWorld-Verified	76.2	4
scicode	59	55.7
simplebench	79.6	80.6
simpleqa_verified	77.3	56.7
SWE-bench Multilingual	76.9	89.5
SWE-bench Pro	54.2	79.2
SWE-bench Verified	80.6	96
TauBench V3 - Banking	21.4	44.7
Terminal-Bench 2.1	74	89.1

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.