Which is better, Claude Sonnet 5 or Gemini 3.1 Pro?

Across 45 shared benchmarks tracked on Vector Wire, Claude Sonnet 5 leads on 27 and Gemini 3.1 Pro leads on 18 — Claude Sonnet 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 Sonnet 5 vs Gemini 3.1 Pro

AnthropicvsGoogle45 shared benchmarks27–18 head-to-head

Benchmark	Claude Sonnet 5	Gemini 3.1 Pro
AA Agentic Index	49.7	23
AA Intelligence	55.3	47.7
AA-LCR	77	79
AA-Omniscience	16.4	32.9
arena_elo	1463	1486
arena_text_factuality	1453	1471
Artificial Analysis Coding Index	71.5	68.8
browsecomp	84.7	85.9
charxiv_reasoning_with_tools	88.3	83.2
charxiv_rq	77	89.9
coding_arena_elo	1540	1447
critpt	16.9	17.7
DeepSWE 1.1	54	12
gdpval	54.8	23.3
GDPval-AA v2	1618	962
GDPVal-AA v2 (Elo)	1598	965
GPQA Diamond	91.1	94.3
HLE	57.4	51.4
Legal Agent Benchmark	5.8	0
livebench_agentic_coding	59.4	44.1
livebench_coding	80.7	76.5
livebench_data_analysis	71.7	78.5
livebench_instruction_following	63.9	79.1
livebench_language	75	85.4
livebench_math	92.9	91
livebench_reasoning	88.7	84
MLE-Bench	66.9	42.6
MMMU-Pro	77.3	83
mrcr_v2_8needle_128k_average	81.5	84.9
officeqa_pro	59.4	42.9
OmniScience Accuracy	40	55.3
OmniScience Non-Hallucination	60.6	50.1
OSWorld-Verified	81.2	76.2
scicode	53.6	59
simplebench	60.6	79.6
simpleqa_verified	25	77.3
SWE-bench Multilingual	78.3	76.9
SWE-bench Pro	63.2	54.2
SWE-bench Verified	85.2	80.6
TauBench V3 - Banking	37.3	21.4
Terminal-Bench 2.0	80.4	68.5
Terminal-Bench 2.1	80.5	74
toolathlon	54.3	48.8
usamo_2026	79.5	74.4
τ³-Bench	28.2	67.1

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.