Compare the listed model data for Qwen3 235B-A22B (MoE) by Qwen and Qwen3.5 397B-A17B (MoE) by Qwen. Benchmark and pricing records are not a controlled test and may change.
Data last updated: 2026-10-16. Verify current model details with each provider.
What the listed data says
Qwen3 235B-A22B (MoE): 128K tokens; Qwen3.5 397B-A17B (MoE): 128K tokens. The listed context-window values are equal. Qwen3 235B-A22B (MoE) is listed for coding, reasoning, agentic workflows, long-context tasks; Qwen3.5 397B-A17B (MoE) is listed for no specific capability flags. Shared catalog benchmark entries: MMLU: Qwen3 235B-A22B (MoE) 93.5%; Qwen3.5 397B-A17B (MoE) 94.1%; LiveCodeBench: Qwen3 235B-A22B (MoE) 70.7%; Qwen3.5 397B-A17B (MoE) 73.5%; AIME 2025: Qwen3 235B-A22B (MoE) 85.7%; Qwen3.5 397B-A17B (MoE) 88.2%. These records are reference data, not a controlled head-to-head test; verify current pricing and specifications with each provider.
Catalog source links
Use these model and provider links to verify current specifications, pricing, and availability.
The first listed parameter quantities are 235B total (22B active via MoE) and 397B total (17B active); the larger quantity is 1.69× the smaller. Total and active MoE parameters are not directly equivalent.
A release-date gap cannot be calculated from the catalog values (Qwen3 235B-A22B (MoE): April 2025; Qwen3.5 397B-A17B (MoE): 2025).
Cost example: 1M input and 1M output tokens
A cost example for 1M input plus 1M output tokens cannot be calculated reliably because separate numeric input and output rates are not listed for both models.
Choose Qwen3 235B-A22B (MoE) if…
Choose Qwen3 235B-A22B (MoE) when your task aligns with catalog-listed uses or strengths such as Frontier-level coding: production systems, open-source contributions, Leetcode hard, Competition mathematics, physics, and chemistry at graduate level, Enterprise AI applications requiring highest available open-model quality, Long-form technical writing: whitepapers, research papers, documentation.
Choose Qwen3.5 397B-A17B (MoE) if…
Choose Qwen3.5 397B-A17B (MoE) when your task aligns with catalog-listed uses or strengths such as Frontier-level AI applications with efficiency constraints, Production deployments of Qwen3.5 quality at scale, Qwen3.5 flagship model — 397B knowledge at 17B inference cost, Near Qwen3 235B quality at more efficient inference.
Feature Comparison — Qwen3 235B-A22B (MoE) vs Qwen3.5 397B-A17B (MoE)
Catalog specifications for Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)
Specification
Qwen3 235B-A22B (MoE)
Qwen3.5 397B-A17B (MoE)
Provider
Qwen
Qwen
Parameters
235B total (22B active via MoE)
397B total (17B active)
Context window
128K tokens
128K tokens
Maximum output
Not listed
Not listed
Catalog pricing
Open source — free to self-host. Third-party API ~$0.30-0.80/MTok
Apache 2.0 — free
Listed input/output cost per 1M tokens
Separate numeric input/output rates not listed
input $0.8, output $1.5 per 1M tokens
License
Apache 2.0
Apache 2.0
Coding support
Listed
Not listed
Reasoning support
Listed
Not listed
Vision input
Not listed
Not listed
Multimodal
Not listed
Not listed
Open weights
Yes
Yes
Benchmark: MMLU
93.5%
94.1%
Benchmark: LiveCodeBench
70.7%
73.5%
Benchmark: AIME 2025
85.7%
88.2%
Figures are transcribed from LITHOVEX model records. Benchmark versions and evaluation setups may differ; pricing and specifications can change. Check each provider's current documentation before making a decision.
Frequently Asked Questions
What is different between Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)?
Qwen3 235B-A22B (MoE): 128K tokens; Qwen3.5 397B-A17B (MoE): 128K tokens. The listed context-window values are equal. Qwen3 235B-A22B (MoE) is listed for coding, reasoning, agentic workflows, long-context tasks; Qwen3.5 397B-A17B (MoE) is listed for no specific capability flags. Shared catalog benchmark entries: MMLU: Qwen3 235B-A22B (MoE) 93.5%; Qwen3.5 397B-A17B (MoE) 94.1%; LiveCodeBench: Qwen3 235B-A22B (MoE) 70.7%; Qwen3.5 397B-A17B (MoE) 73.5%; AIME 2025: Qwen3 235B-A22B (MoE) 85.7%; Qwen3.5 397B-A17B (MoE) 88.2%. These records are reference data, not a controlled head-to-head test; verify current pricing and specifications with each provider.
What benchmark values are listed for Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)?
MMLU: Qwen3 235B-A22B (MoE) 93.5%; Qwen3.5 397B-A17B (MoE) 94.1%; LiveCodeBench: Qwen3 235B-A22B (MoE) 70.7%; Qwen3.5 397B-A17B (MoE) 73.5%; AIME 2025: Qwen3 235B-A22B (MoE) 85.7%; Qwen3.5 397B-A17B (MoE) 88.2%. The catalog values may use different versions and evaluation setups, so they are reference points rather than a controlled head-to-head result.
Which model has the larger listed context window?
Qwen3 235B-A22B (MoE): 128K tokens; Qwen3.5 397B-A17B (MoE): 128K tokens. The listed context-window values are equal.
What are the listed costs per million tokens for Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)?
Qwen3 235B-A22B (MoE): Separate numeric input/output rates not listed (Open source — free to self-host. Third-party API ~$0.30-0.80/MTok). Qwen3.5 397B-A17B (MoE): input $0.8, output $1.5 per 1M tokens (Apache 2.0 — free). Where separate numeric input and output rates are available, the worked example uses 1M input plus 1M output tokens. Confirm current terms with each provider.
How far apart are the listed release dates for Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)?
The catalog lists Qwen3 235B-A22B (MoE) as April 2025 and Qwen3.5 397B-A17B (MoE) as 2025; a release-date gap cannot be calculated from both records.
When was the comparison data last updated?
The model catalog records 2026-10-16 as its data update date. Provider specifications and prices may have changed since then.
How should I choose between Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE)?
Compare the listed specifications and capabilities, review the benchmark notes, then test both with the same representative prompt for your task. This page does not claim a universal winner.
Create your free LITHOVEX account and compare Qwen3 235B-A22B (MoE) and Qwen3.5 397B-A17B (MoE) on your actual tasks. No credit card, no API key, no subscriptions needed.