Which is better, GPT-5.2 or Opus 5?

Across 36 shared benchmarks tracked on Vector Wire, GPT-5.2 leads on 3 and Opus 5 leads on 33 — 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 —

GPT-5.2 vs Opus 5

OpenAIvsAnthropic36 shared benchmarks3–33 head-to-head

Benchmark	GPT-5.2	Opus 5
AA Agentic Index	60.2	59.2
AA Intelligence	43.3	63.1
AA-LCR	79.3	78.7
AA-Omniscience	0.3	37.1
arc_agi_1	86.2	97.5
ARC-AGI-2	52.9	90.4
arena_elo	1437	1487
arena_text_factuality	1442	1481
Artificial Analysis Coding Index	48.7	78
browsecomp	65.8	90.8
coding_arena_elo	1418	1662
critpt	11.6	29.1
gdpval	48.3	67.2
GDPval-AA (Elo)	1462	1827
GPQA Diamond	92.4	93.7
healthbench_length_adjusted	56.8	57.8
HLE	45.5	64.7
HLE (with tools)	45.5	64.7
livebench_agentic_coding	50.3	65.2
livebench_coding	76.1	81.5
livebench_data_analysis	78.2	74.5
livebench_instruction_following	61.8	63.8
livebench_language	79.8	88.7
livebench_math	93.2	95.7
livebench_reasoning	83.2	91.2
MCP Atlas	68	85.8
MMMU-Pro	79.5	84.7
OmniScience Accuracy	44.3	60.9
OmniScience Non-Hallucination	38.4	40.5
scicode	52.1	55.7
simplebench	45.8	80.6
simpleqa_verified	38.9	56.7
SWE-bench Multilingual	72	89.5
SWE-bench Pro	55.6	79.2
SWE-bench Verified	80	96
TauBench V3 - Banking	11.1	44.7

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.