Which is better, Claude Opus 4.5 or Qwen3.7 Max?

Across 47 shared benchmarks tracked on Vector Wire, Claude Opus 4.5 leads on 15 and Qwen3.7 Max leads on 32 — Qwen3.7 Max 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 4.5 vs Qwen3.7 Max

AnthropicvsAlibaba47 shared benchmarks15–32 head-to-head

Benchmark	Claude Opus 4.5	Qwen3.7 Max
AA Agentic Index	59.6	30.9
AA Intelligence	41.9	46.7
AA-LCR	76	74.7
AA-Omniscience	14	13.5
aime_2026	95.1	97
arena_elo	1473	1474
arena_text_factuality	1458	1479
Artificial Analysis Coding Index	47.8	66
Claw Eval (pass@3)	59.6	65.2
coding_arena_elo	1495	1517
critpt	4.6	13.4
gdpval	47.3	38.5
GPQA Diamond	87	92.4
HLE	30.8	41.4
HLE (with tools)	43.4	53.5
hmmt_feb_2026	85.3	97.1
hmmt_nov_2025	93.3	95
IFBench	58	80.5
imo_answer_bench	84	90
livebench	76	74.3
livebench_agentic_coding	39.7	43.6
livebench_coding	79.7	74.2
livebench_data_analysis	74.4	71.8
livebench_instruction_following	62.5	74
livebench_language	81.3	79.7
livebench_math	90.4	85.3
livebench_reasoning	80.1	83.3
LiveCodeBench v6	84.8	91.6
longbench_v2	64.4	65.3
MCP Atlas	69.8	76.4
mmlu_redux	95.6	95
MMLU-Pro	90	89.6
mmmlu	90.8	90.3
nl2repo	43.2	47.2
OmniScience Accuracy	46.6	31.1
OmniScience Non-Hallucination	39	74.4
PaperBench	72.9	64.8
scicode	50	53.5
simplebench	62	70.4
simpleqa_verified	41.8	58.5
supergpqa	70.6	73.6
SWE-bench Multilingual	77.5	78.3
SWE-bench Pro	57.1	60.6
SWE-bench Verified	81.5	80.4
Terminal-Bench 2.0	59.3	69.7
Terminal-Bench Hard	47	50.8
τ²-Bench Telecom (AA run)	89.5	94.7

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.