Model comparison · accepted catalogue facts 2026-11-22

Qwen3 235B A22B Thinking 2507 vs Qwen3.5 397B A17B

Qwen3.5 397B A17B has an equal-or-higher measured score and an equal-or-lower estimated price. No known model-capability loss from Qwen3 235B A22B Thinking 2507; endpoint and task checks still apply. The reported score intervals do not overlap; this is not a task-specific success guarantee.

Both models belong to the Alibaba Qwen family. Inclusion does not recommend either model. This is a direct pair comparison; global frontier membership is not its verdict.

Qwen3 235B A22B Thinking 2507

LMArena general · Elo · higher is better

Scale starts at 1040 Elo

1415.4

Balanced effective USD / 1M tokens · lower is better

$1.21/M

Qwen3.5 397B A17B

LMArena general · Elo · higher is better

Scale starts at 1040 Elo

1437.6

Balanced effective USD / 1M tokens · lower is better

$0.878/M

In-Depth Analysis

Qwen3 235B A22B Thinking 2507 vs Qwen3.5 397B A17B: Measured Technical Breakdown

Comprehensive comparison across measured human-preference evaluation, effective production token costs, architectural limits, and migration trade-offs.

Measured Capability (LMArena)
Qwen3.5 397B A17B leads by 22.2 Elo

Qwen3 235B A22B Thinking 2507: 1415.4 ±6.53 · Qwen3.5 397B A17B: 1437.6 ±3.03 Statistically significant preference separation

Cost Comparison Across Canonical Production Workloads

Rates represent effective cost per 1M tokens under standard token blends. Cached prompt reads are factored according to workload context reuse.

Workload & RatioQwen3 235B A22B Thinking 2507Qwen3.5 397B A17BCost DifferenceAdvantage
Balanced 3 tokens in per 1 out$1.21 /1M tokens$0.878 /1M tokens$0.335 (28% diff)Qwen3.5 397B A17B costs less
Summarise Huge input, tiny output$0.603 /1M tokens$0.488 /1M tokens$0.115 (19% diff)Qwen3.5 397B A17B costs less
Chat Long cached system prompt$1.67 /1M tokens$1.17 /1M tokens$0.500 (30% diff)Qwen3.5 397B A17B costs less
Code gen Output dominates$2.28 /1M tokens$1.56 /1M tokens$0.720 (32% diff)Qwen3.5 397B A17B costs less
Agentic Context replayed each turn$0.908 /1M tokens$0.683 /1M tokens$0.225 (25% diff)Qwen3.5 397B A17B costs less

Architectural & Operational Specifications

SpecificationQwen3 235B A22B Thinking 2507Qwen3.5 397B A17BOperational Note
Context window128,000 tokens262,144 tokensQwen3.5 397B A17B has a 2.0× larger context window
Max output generation16,384 tokens81,920 tokensQwen3.5 397B A17B allows longer single-turn completions
Base input price$0.450 / 1M tokens$0.390 / 1M tokensQwen3.5 397B A17B input rate is cheaper
Base output price$3.50 / 1M tokens$2.34 / 1M tokensQwen3.5 397B A17B output rate is cheaper
Cached input priceNot listed / standard rateNot listed / standard rate—
Reasoning & thinking tokensSupported; standard ratesSupported; standard rates—
Input modalitiestexttext, image, video—
Tool calling / function useSupportedSupported—
Open weights & licenceProprietary (API-only)Proprietary (API-only)—
Provider & reseller marketQwen · 1 provider offersQwen · 8 provider offers (2.08× spread)—

When to Choose Qwen3 235B A22B Thinking 2507

  • Direct integration with Qwen tooling and API features.
Risks when moving from Qwen3.5 397B A17B → Qwen3 235B A22B Thinking 2507:
  • Context window: 262,144 → 128,000 tokens.
  • Output limit: 81,920 → 16,384 tokens.
  • image input is not supported by the candidate.
  • video input is not supported by the candidate.

When to Choose Qwen3.5 397B A17B

  • Higher measured general capability (1437.6 Elo on human evaluation).
  • Larger context window (262,144 tokens) for long-document or repo-level context.
  • Higher maximum completion limit (81,920 tokens).
  • Broader multimodal input support: text, image, video.
  • Direct integration with Qwen tooling and API features.

Compare · model decision

Choose two or three models

Inspect the differences, estimate your workload and keep the evidence. All standard catalogue models are available, including unrated ones.

Catalogue facts
2026-11-22

2 / 3 selected · 0 requirements
Open public comparison link
Public workload and requirements

These settings are included in the share link. The fixed-request scenario and private evaluation data below are excluded.

A ↔ B

Qwen3.5 397B A17B has an equal-or-higher measured score and an equal-or-lower estimated price. No known model-capability loss from Qwen3 235B A22B Thinking 2507; endpoint and task checks still apply. The reported score intervals do not overlap; this is not a task-specific success guarantee.

Estimated blend: 75% input, 25% output; 0% cached input. Base context tier. Prompt length selects the billing tier. An unlisted cache-read rate uses that tier’s full input rate in this blend; missing required input or output rates prevent an estimate. The fixed-request estimate below validates actual context and output limits and requires a published cache-read rate when cache hits are assumed.

What changes between these models

Quality first, then estimated price. LMArena scores are human-preference measurements, not a benchmark of your application. A missing score is unrated.

  • Qwen3 235B A22B Thinking 2507 · Meets the stated model requirements
  • Qwen3.5 397B A17B · Meets the stated model requirements
Material differences and unknown evidence. Deltas compare each model with model A. Unknown facts remain visible.
Measured factA Qwen3 235B A22B Thinking 2507B Qwen3.5 397B A17B
LMArena general score (Elo)1,415.41,437.6Δ vs A: +22.2 Elo
Score interval half-width (± Elo)6.533.03Δ vs A: -3.5 Elo
LMArena document score (Elo)UnknownUnknown
Estimated effective USD / 1M tokens$1.2125$0.8775Δ vs A: -0.335 USD/M
Base input USD / 1M tokens$0.45$0.39Δ vs A: -0.06 USD/M
Base output USD / 1M tokens$3.5$2.34Δ vs A: -1.16 USD/M
Base cached-input USD / 1M tokensUnknownUnknown
Context window (tokens)128,000262,144Δ vs A: +134,144 tokens
Output limit (tokens)16,38481,920Δ vs A: +65,536 tokens
Input modalitiestexttext, image, video
Open weightsUnknownUnknown
Recorded licenceUnknownUnknown
Distinct sellers in the catalogue16Δ vs A: +5
Serving precision differs across offersNot supportedSupportedΔ vs A: +1
Recorded retirement dateUnknownUnknown
5 shared measured facts
Shared measured facts. Deltas compare each model with model A. Unknown facts remain visible.
Measured factA Qwen3 235B A22B Thinking 2507B Qwen3.5 397B A17B
Benchmark reasoning effortdefaultdefault
Tool useSupportedSupportedΔ vs A: 0
ReasoningSupportedSupportedΔ vs A: 0
Model providerQwenQwen
Retirement evidenceNone announced; not a guaranteeNone announced; not a guarantee

Provider counts and precision flags describe catalogue offers. They do not establish the precision, availability, region, tool behaviour or limits of your chosen endpoint. Open each model’s provider table before switching.

Complete price ladders and billing conditions

Qwen3 235B A22B Thinking 2507

Rates below are USD per 1M tokens. A threshold selects that whole input/output rate pair.

Input-token thresholdInputOutputCached input
Base, up to and including first threshold$0.45$3.5Unknown

Qwen3.5 397B A17B

Rates below are USD per 1M tokens. A threshold selects that whole input/output rate pair.

Input-token thresholdInputOutputCached input
Base, up to and including first threshold$0.39$2.34Unknown
Migration preflight · compare each candidate with model A

Keeping Qwen3 235B A22B Thinking 2507 is a valid decision. No pairwise comparison certifies a drop-in replacement.

Qwen3 235B A22B Thinking 2507 → Qwen3.5 397B A17B

No known model-capability loss

  • Provider endpoint, serving precision, availability and task success need your own verification. Catalogue model capabilities do not certify an endpoint.
Check alternatives to Qwen3 235B A22B Thinking 2507

Estimate a fixed request shape

Private to this tab unless you save locally or explicitly export private details. These are estimates, not observed invoices. Cross-model token counts and task success are assumptions.

0.1 extra attempts means 10 extra full attempts per 100 requests. Cache writes, tool charges, taxes and unlisted charges are excluded.

  • Qwen3 235B A22B Thinking 2507Estimated total: $2.65 1,000 full attempts · base rates
  • Qwen3.5 397B A17BEstimated total: $1.95 1,000 full attempts · base rates
Prompt and output sensitivity · sampled scenarios

Every point recalculates the entire request shape. Tier-boundary samples include one token below, at and above each threshold. Ordering between samples is not guaranteed. Lowest known cost excludes unknown estimates and is not a replacement recommendation.

Vary input; hold output, requests, cache and retries fixed.
Input tokensQwen3 235B A22B Thinking 2507Qwen3.5 397B A17BCoverage
2,000$2.65$1.95Lowest knownAll computed
4,096$3.5932$2.76744Lowest knownAll computed
32,768$16.4956$13.94952Lowest knownAll computed
131,072ineligibleInput plus output exceeds the published context window.$52.28808Lowest knownIncomplete
Vary output; hold input, requests, cache and retries fixed.
Output tokensQwen3 235B A22B Thinking 2507Qwen3.5 397B A17B
128$1.348$1.07952
2,048$8.068$5.57232
8,192$29.572$19.94928

Bring your observed evaluation results

JSON is parsed locally. Prompts, filenames, individual responses and extra fields are not retained. Import only aggregate measurements for selected models. Task/version labels do not establish that two evaluations used the same dataset.

Expected format
[{"model":"provider/exact-model-id","task":"support-answer","version":"rubric-1","attempts":100,"successes":84,"costUsd":2.4,"latencyMs":950}]

Use the exact catalogue ID. Replace the example values with observed measurements. Omit costUsd or latencyMs when they are unknown. No public score is inferred from these observations.

Keep the decision and its evidence

A portable record retains selected facts, complete tier ladders, public requirements and the publication manifest. It also embeds the full permitted public catalogue to verify the input artifact checksum. Files are larger for this reason (up to 2 MB); local storage is limited to 12 records. Team review is local to this browser; there is no hosted collaboration.

Saving is opt-in and includes private review fields. Shared comparison links never include them. Version 1 Check references remain separate and are not overwritten or guessed into this format.

Saved local decisions · 0

No version 2 decision records are stored in this browser.

Questions this page answers

Is Qwen3 235B A22B Thinking 2507 a better-and-cheaper replacement for Qwen3.5 397B A17B?

Qwen3.5 397B A17B has an equal-or-higher measured score and an equal-or-lower estimated price. No known model-capability loss from Qwen3 235B A22B Thinking 2507; endpoint and task checks still apply. The reported score intervals do not overlap; this is not a task-specific success guarantee.

Which model is higher rated: Qwen3 235B A22B Thinking 2507 or Qwen3.5 397B A17B?

Qwen3.5 397B A17B scores higher on LMArena general (1437.6 ±3.0) compared to Qwen3 235B A22B Thinking 2507 (1415.4 ±6.5), a margin of 22.2 Elo points. The measured confidence intervals do not overlap, demonstrating a statistically significant preference lead on human evaluations.

Which model is more cost-effective: Qwen3 235B A22B Thinking 2507 or Qwen3.5 397B A17B?

On a standard balanced workload (3 prompt tokens per 1 generated token), Qwen3.5 397B A17B costs $0.878 per 1M tokens versus $1.21 for Qwen3 235B A22B Thinking 2507. Qwen3 235B A22B Thinking 2507 is approximately 1.4× the price of Qwen3.5 397B A17B. Base input rates are $0.450 for Qwen3 235B A22B Thinking 2507 and $0.390 for Qwen3.5 397B A17B; base output rates are $3.50 for Qwen3 235B A22B Thinking 2507 and $2.34 for Qwen3.5 397B A17B.

How do their context windows and generation limits compare?

Qwen3 235B A22B Thinking 2507 provides a 128,000-token context window with a maximum output limit of 16,384 tokens. Qwen3.5 397B A17B provides a 262,144-token context window with a maximum output of 81,920 tokens. Qwen3.5 397B A17B offers a 2.0× larger context capacity, making it suitable for larger repositories and multi-document workflows.

What capabilities are lost when switching between Qwen3 235B A22B Thinking 2507 and Qwen3.5 397B A17B?

Switching from Qwen3 235B A22B Thinking 2507 to Qwen3.5 397B A17B causes no detected architectural capability loss. Conversely, moving from Qwen3.5 397B A17B to Qwen3 235B A22B Thinking 2507 involves: Context window: 262,144 → 128,000 tokens; Output limit: 81,920 → 16,384 tokens; image input is not supported by the candidate; video input is not supported by the candidate. Run test suites on your specific prompt templates and tool declarations before cutting over production traffic.

Can Qwen3 235B A22B Thinking 2507 or Qwen3.5 397B A17B be self-hosted or deployed privately?

Both Qwen3 235B A22B Thinking 2507 and Qwen3.5 397B A17B are proprietary cloud models offered via managed API endpoints. Qwen3 235B A22B Thinking 2507 is served by Qwen and 1 catalogue offers, while Qwen3.5 397B A17B is served by Qwen and 8 catalogue offers.

What is included in this comparison?

Accepted model capabilities, context and output limits, complete price-tier ladders, independent LMArena measurements and missing evidence. Provider endpoint precision and application success still need verification.

Can I compare a different workload or keep a decision record?

Yes. Change the public workload, prompt length and requirements in the comparison workspace. Fixed-request scenarios, aggregate evaluations and review notes stay local unless explicitly exported.

Evidence & Ask