Which is better, Claude Opus 4.6 or DeepSeek-V4-Pro?

Across 67 shared benchmarks tracked on Vector Wire, Claude Opus 4.6 leads on 48 and DeepSeek-V4-Pro leads on 18 — Claude Opus 4.6 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 4.6 vs DeepSeek-V4-Pro

AnthropicvsDeepSeek67 shared benchmarks48–18 head-to-head

Benchmark	Claude Opus 4.6	DeepSeek-V4-Pro
AA Agentic Index	67.6	63.3
AA Intelligence	44.9	53
AA-LCR	74.3	70
AA-Omniscience	13.7	-10.6
AgentWorldBench - Android	61.7	55.2
AgentWorldBench - MCP	69.9	63.3
AgentWorldBench - OS	70.2	63.7
AgentWorldBench - Overall	57.8	53
AgentWorldBench - Search	29.3	27.6
AgentWorldBench - SWE	64.5	59.4
AgentWorldBench - Terminal	57.5	51.3
AgentWorldBench - Web	51.4	50.3
aime_2026	96.7	96.7
Apex (Pass@1)	34.5	38.3
Apex Shortlist (Pass@1)	85.9	90.2
apexAgents	33	24.3
arena_elo	1497	1457
arena_text_factuality	1487	1450
Artificial Analysis Coding Index	48.1	59.4
browsecomp	86.8	83.4
Chinese SimpleQA (C-SimpleQA)	76.4	77.7
coding_arena_elo	1545	1582
CorpusQA 1M (ACC)	71.7	62
critpt	12.6	13
cybergym	73.8	52.7
gdpval	55.9	49
GDPval-AA (Elo)	1619	1554
GPQA Diamond	91.3	90.5
HLE	53.1	48.2
HLE (with tools)	62.7	48.2
hmmt_feb_2026	96.2	95.2
hmmt_nov_2025	96.3	94.4
IFBench	62.5	76.5
imo_answer_bench	75.3	89.8
livebench	76.3	73.6
livebench_agentic_coding	49	42.6
livebench_coding	78.2	70
livebench_data_analysis	69.9	74.5
livebench_instruction_following	63.3	62.4
livebench_language	83.3	78.1
livebench_math	89.3	90.7
livebench_reasoning	88.7	82.7
LiveCodeBench	88.8	93.5
MCP Atlas	76.8	74.2
MCPAtlas Public (Pass@1)	73.8	73.6
MMLU-Pro	89.1	87.5
MRCR 1M (MMR)	92.9	83.5
nl2repo	49.8	38.5
OmniScience Accuracy	47	43
OmniScience Non-Hallucination	37.2	12.2
scicode	52	50
simplebench	67.6	50.9
simpleqa_verified	46.5	57.9
SWE-bench Multilingual	77.8	76.2
SWE-bench Pro	57.3	55.4
SWE-bench Verified	80.8	80.6
Terminal-Bench 2.0	65.4	67.9
Terminal-Bench Hard	65.4	46.2
tool_decathlon	47.2	52.8
toolathlon	56.8	51.8
usamo_2026	66.2	60.7
vectara_answer_rate	99.8	97.2
vectara_avg_summary_length	137.6	153.8
vectara_factual_consistency	87.8	91.4
vectara_hallucination_rate ↓	12.2	8.6
τ²-Bench Telecom (AA run)	99.3	96.2
τ³-Bench	72.4	25.8

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.