Which is better, Claude Opus 4.6 or Qwen3.5 35B A3B?

Across 57 shared benchmarks tracked on Vector Wire, Claude Opus 4.6 leads on 42 and Qwen3.5 35B A3B leads on 14 — 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 Qwen3.5 35B A3B

AnthropicvsAlibaba57 shared benchmarks42–14 head-to-head

Benchmark	Claude Opus 4.6	Qwen3.5 35B A3B
AA Agentic Index	67.6	44.1
AA Intelligence	44.9	29.9
AA-LCR	74.3	68.3
AA-Omniscience	13.7	-48.1
AgentWorldBench - Android	61.7	53.2
AgentWorldBench - MCP	69.9	57.9
AgentWorldBench - OS	70.2	56.3
AgentWorldBench - Overall	57.8	47.7
AgentWorldBench - Search	29.3	26
AgentWorldBench - SWE	64.5	47.6
AgentWorldBench - Terminal	57.5	46.1
AgentWorldBench - Web	51.4	47.1
aime_2026	96.7	93.3
Artificial Analysis Coding Index	48.1	37
baby_vision	14.8	38.4
browsecomp	86.8	61
charxiv_rq	69.1	77.5
coding_arena_elo	1545	1250
critpt	12.6	0.9
ERQA	40.8	64.8
gdpval	55.9	20.3
GPQA Diamond	91.3	84.5
GSM8K	0.6	90.1
HLE	53.1	47.4
hmmt_feb_2026	96.2	81.8
hmmt_nov_2025	96.3	89.2
IFBench	62.5	72.5
imo_answer_bench	75.3	76.8
LiveCodeBench v6	88.8	74.6
mathvision	71.2	83.9
MCP Atlas	76.8	62.4
mcpmark	56.7	27
MMLU-Pro	89.1	85.3
mmmlu	91.1	85.2
MMMU-Pro	77.3	75.1
multichallenge	56	60
nl2repo	49.8	20.5
ocrbench_v2	48.4	65.3
OmniDocBench 1.5	86.6	89.3
OmniScience Accuracy	47	20.1
OmniScience Non-Hallucination	37.2	14.6
OSWorld-Verified	72.7	54.5
RealWorldQA	73.9	84.1
scicode	52	37.7
screenspot_pro_no_tools	57.7	68.6
SWE-bench Multilingual	77.8	60.3
SWE-bench Pro	57.3	44.6
SWE-bench Verified	80.8	70
Terminal-Bench 2.0	65.4	40.5
Terminal-Bench Hard	65.4	26.5
tool_decathlon	47.2	28.7
vectara_answer_rate	99.8	99.8
vectara_avg_summary_length	137.6	94.9
vectara_factual_consistency	87.8	89.5
vectara_hallucination_rate ↓	12.2	10.5
τ²-Bench Telecom (AA run)	99.3	89.2
τ³-Bench	72.4	68.9

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.