Qwen3 235B A22B 2507 (Reasoning) vs Qwen3.5 397B A17B (Reasoning)

Which is better for the money: Qwen3 235B A22B 2507 (Reasoning) by Alibaba or Qwen3.5 397B A17B (Reasoning) by Alibaba? Intelligence, total benchmark cost, and end-to-end time from Artificial Analysis & other sources, snapshot 2026-10-26.

Trade-off

AlibabaQwen3 235B A22B 2507 (Reasoning)Off the frontier

Intelligence: 12.70
Total cost: $234

AlibabaQwen3.5 397B A17B (Reasoning)Off the frontier

Intelligence: 18.40
Total cost: $941

Verdict

Depends on your budget

Qwen3.5 397B A17B (Reasoning) scores 5.7 points higher (18.4 vs 12.7) but costs 4.0× as much to run the benchmark ($941 vs $234; premium vs mid-range cost classes). They are not alternatives at one budget: pick Qwen3.5 397B A17B (Reasoning) when you need intelligence above 12.7, and Qwen3 235B A22B 2507 (Reasoning) when 12.7 is enough.

Metric	Qwen3 235B A22B 2507 (Reasoning)	Qwen3.5 397B A17B (Reasoning)	Difference
Intelligence (higher is better)	12.7	18.4	+44.9% Qwen3.5 397B A17B (Re… is 44.9% more intelligent
Total AA cost (lower is better)	$233.56	$940.97	+302.9% Qwen3.5 397B A17B (Re… is 4.0× more expensive
End-to-end time (lower is better)	22.73s	65.86s	+189.7% Qwen3.5 397B A17B (Re… is 189.7% slower
Price per 1M tokens (in / out)	$0.23 / $2.30	$0.60 / $3.60
Input modalities	Text	Text, Image

Qwen3.5 397B A17B (Reasoning) vs Qwen3 235B A22B 2507 (Reasoning): 44.9% more intelligence, 302.9% more cost to run the AA benchmark, and 189.7% slower end-to-end.

Open this matchup in the interactive picker Qwen3 235B A22B 2507 (Reasoning): what beats it Qwen3.5 397B A17B (Reasoning): what beats it

Today's Pareto frontier

Against the frontier

Qwen3 235B A22B 2507 (Reasoning) is off the frontier: at the same or lower cost, MiMo-V2.6-Pro scores 46.30 for $207.
Qwen3.5 397B A17B (Reasoning) is off the frontier: at the same or lower cost, GPT-6.1 Sol (Xhigh) scores 51.00 for $662.

Qwen3 235B A22B 2507 (Reasoning)Qwen3.5 397B A17B (Reasoning)Today's Pareto frontierOpen-weights frontierEvery other model

Matchup

Compare something else

Agents: Accept: text/markdown or ?format=md returns this page as Markdown. JSON: /api/v1/compare/qwen3-235b-a22b-instruct-2507-reasoning/qwen3-5-397b-a17b · Markdown: /api/v1/compare/qwen3-235b-a22b-instruct-2507-reasoning/qwen3-5-397b-a17b?format=md · llms.txt · agent skill