Which is better, Claude Sonnet 4.6 or Qwen3.6 Plus?

Across 49 shared benchmarks tracked on Vector Wire, Claude Sonnet 4.6 leads on 37 and Qwen3.6 Plus leads on 11 — Claude Sonnet 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 Sonnet 4.6 vs Qwen3.6 Plus

AnthropicvsAlibaba49 shared benchmarks37–11 head-to-head

Benchmark	Claude Sonnet 4.6	Qwen3.6 Plus
AA Agentic Index	61.6	29
AA Intelligence	48.4	40.5
AA-LCR	74	72.3
AA-Omniscience	12.2	2.6
AgentWorldBench - Android	58	57.6
AgentWorldBench - MCP	70	55.3
AgentWorldBench - OS	63.2	60.3
AgentWorldBench - Overall	56	50.8
AgentWorldBench - Search	28.8	21.9
AgentWorldBench - SWE	64.5	59.1
AgentWorldBench - Terminal	57	50.6
AgentWorldBench - Web	50.8	50.8
arena_elo	1472	1444
arena_text_factuality	1460	1438
Artificial Analysis Coding Index	63	54.5
coding_arena_elo	1523	1460
critpt	3.1	2.9
Finance Agent v2	51	40.9
frontiermath_tier_4	8.3	8.3
gdpval	54.8	32
GPQA Diamond	89.9	90.4
HLE	49	28.8
HLE (with tools)	46.8	50.6
IFBench	56.6	75.2
livebench	75.5	70.8
livebench_agentic_coding	42.6	41.4
livebench_coding	79.3	78.2
livebench_data_analysis	78	69.9
livebench_instruction_following	63.2	58.3
livebench_language	76.1	75
livebench_math	87	83.7
livebench_reasoning	84.8	75.8
MCP Atlas	69.5	74.1
mmmlu	89.3	89.5
MMMU-Pro	75.6	78.8
OmniScience Accuracy	40.9	26.4
OmniScience Non-Hallucination	51.6	68
OSWorld-Verified	78.5	62.5
scicode	47	40.7
simpleqa_verified	29	49.1
SWE-bench Pro	58.1	56.6
SWE-bench Verified	79.6	78.8
TauBench V3 - Banking	34.4	20.8
Terminal-Bench 2.0	59.1	61.6
Terminal-Bench 2.1	71.2	61.4
Terminal-Bench Hard	59.1	43.9
toolathlon	49.4	39.8
τ²-Bench Telecom (AA run)	97.9	97.7
τ³-Bench	30.5	70.7

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.