Which is better, Claude Opus 4.7 or GPT-5.6 Sol?

Across 46 shared benchmarks tracked on Vector Wire, Claude Opus 4.7 leads on 10 and GPT-5.6 Sol leads on 36 — GPT-5.6 Sol 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 GPT-5.6 Sol

AnthropicvsOpenAI46 shared benchmarks10–36 head-to-head

Benchmark	Claude Opus 4.7	GPT-5.6 Sol
AA Agentic Index	64.6	57.8
AA Intelligence	55	61
AA-LCR	75.3	77.7
AA-Omniscience	27.3	22
aime_2026	95.8	99.9
arc_agi_1	92	97.5
arc_agi_3	0.2	7.8
ARC-AGI-2	75.8	92.5
arena_elo	1494	1482
arena_text_factuality	1474	1474
Artificial Analysis Coding Index	73.6	78.3
browsecomp	79.8	90.4
charxiv_rq	82.1	87.8
coding_arena_elo	1557	1619
critpt	12	32.3
cybergym	73.8	83.6
Finance Agent v2	51.5	53.8
gdpval	58.6	61.1
GPQA Diamond	94.2	94.6
HiL-Bench	41.7	32.3
HLE	54.7	49.5
HLE (with tools)	54.7	58
IFBench	58.6	72.7
itbenchSre	46.7	56.2
livebench_agentic_coding	50.7	56.2
livebench_coding	82.1	83.9
livebench_data_analysis	78.3	79.8
livebench_instruction_following	66.7	71.8
livebench_language	77.9	87.7
livebench_math	92.8	96.2
livebench_reasoning	87.2	91.7
MCP Atlas	79.1	83.6
MMMU-Pro	78.8	84.6
officeqa_pro	80.6	63.2
OmniScience Accuracy	48.9	59.4
OmniScience Non-Hallucination	57.7	10.6
OSWorld-Verified	82.8	83
scicode	54.5	56.9
simplebench	61.7	64.8
simpleqa_verified	50.6	71.6
SWE-bench Pro	64.3	64.6
TauBench V3 - Banking	34.6	44.3
Terminal-Bench 2.1	83.1	89.5
Terminal-Bench Hard	54.5	65.9
toolathlon	59.3	58
τ²-Bench Telecom (AA run)	88.6	85.1

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.