Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

Start free brief
BenchLM
Model A
Qwen3.5 397B

Alibaba

57.5/100

Estimated · Public rank #82

90% interval 45.9–69.0

Qwen3.5 397B vs Qwen3.6 Plus

Updated August 15, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Model B
Qwen3.6 Plus

Alibaba

64.8/100

Supported · Public rank #36

90% interval 55.7–73.9

Decision reading

Qwen3.6 Plus has the higher public score estimate, 64.79 versus 57.46, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

37 results are shared. Category rows based on different benchmark sets are marked directional and do not name a winner.

Which one for your work

Recommendations appear only when a shared evidence basis or an explicit operating constraint supports the call. Secondary and unsupported use cases stay disclosed below the initial list.

  • Coding work

    Code generation, repair, and software-engineering tasks

    Qwen3.6 Plus

    Qwen3.6 Plus leads on the same 2 weighted benchmark rows.

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    Qwen3.6 Plus

    Qwen3.6 Plus has the larger documented context window.

    Confidence: documented

Show secondary and unsupported calls

  • Agentic work

    Tool use, computer use, and multi-step task completion

    Not enough matched evidence

    The category averages use different weighted benchmark sets, so they are directional rather than like-for-like.

    Confidence: limited

  • Chat turn cost

    1K fresh input + 500 output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

  • Cache-heavy agent loop cost

    200K cached + 20K fresh input + 10K output tokens

    Not enough matched evidence

    The page does not recommend a cost winner because at least one model cannot fit the stated workload in one request. Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate. Qwen3.6 Plus has no comparable published API token rate.

    Confidence: rate-fallback

  • Repository review cost

    50K fresh input + 3K output tokens

    Not enough matched evidence

    A complete comparable API-rate estimate is not available for both models.

    Confidence: listed-rates

What is actually comparable

Shared results can support a head-to-head reading. Results present for only one model describe coverage, not superiority.

Shared results
37
Qwen3.5 397B only
1
Qwen3.6 Plus only
7
Like-for-like categories
5 / 8

3 categories use different evidence sets. Those rows remain visible for coverage context but do not name a winner.

Category results, on a stated basis

Each row states whether both averages use the same weighted benchmark set. Directional and not-comparable rows remain visible, but they never receive a winner in this template.

Coding

Like-for-like
Qwen3.5 397B
66.5
Qwen3.6 Plus
70.3
Weighted basis
2 vs 2 rows
Reading
Qwen3.6 Plus leads

Reasoning

Like-for-like
Qwen3.5 397B
63.2
Qwen3.6 Plus
62.0
Weighted basis
1 vs 1 rows
Reading
Qwen3.5 397B leads

Knowledge

Like-for-like
Qwen3.5 397B
56.6
Qwen3.6 Plus
57.1
Weighted basis
4 vs 4 rows
Reading
Qwen3.6 Plus leads

Multilingual

Like-for-like
Qwen3.5 397B
84.7
Qwen3.6 Plus
84.7
Weighted basis
1 vs 1 rows
Reading
Tie

Multimodal

Like-for-like
Qwen3.5 397B
79.6
Qwen3.6 Plus
79.8
Weighted basis
2 vs 2 rows
Reading
Qwen3.6 Plus leads

Agentic

Directional only
Qwen3.5 397B
56.5
Qwen3.6 Plus
61.6
Weighted basis
2 vs 1 rows
Reading
Directional only

Math

Directional only
Qwen3.5 397B
90.6
Qwen3.6 Plus
60.5
Weighted basis
2 vs 4 rows
Reading
Directional only

Instruction following

Directional only
Qwen3.5 397B
92.6
Qwen3.6 Plus
82.3
Weighted basis
1 vs 2 rows
Reading
Directional only

Shape of the matched evidence

Only shared public evidence is shown. Sparse evidence stays a ruled list rather than being closed into a radar shape.

Too few matched category axes support a radar. The ruled list below shows only shared benchmark results; positions use each benchmark’s normalized display scale when available.

  • Terminal-Bench 2.0

    Agentic

    Qwen3.5 397B: 52.5%Qwen3.6 Plus: 61.6%Normalized gap 9.1Shared source

  • SWE-bench Pro

    Coding

    Qwen3.5 397B: 50.9%Qwen3.6 Plus: 56.6%Normalized gap 5.7Shared source

  • SWE-bench Verified

    Coding

    Qwen3.5 397B: 76.2%Qwen3.6 Plus: 78.8%Normalized gap 2.6Shared source

  • GPQA

    Knowledge

    Qwen3.5 397B: 88.4%Qwen3.6 Plus: 90.4%Normalized gap 2.0Shared source

  • AIME26

    Math

    Qwen3.5 397B: 93.3%Qwen3.6 Plus: 95.3%Normalized gap 2.0Shared source

What each workload costs

Three fixed token mixes turn per-token rates into comparable decisions. Each scenario states context fit and whether cached input had to fall back to the published list-input rate.

Chat turn

1K fresh input + 500 output tokens

Qwen3.5 397B
$0.0024
Fits in one request
Qwen3.6 Plus
API rate not published
Fits in one request

Qwen3.6 Plus has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Qwen3.5 397B
$0.0408
Fits in one request
Qwen3.6 Plus
API rate not published
Fits in one request

Qwen3.6 Plus has no comparable published API token rate.

Cache-heavy agent loop

200K cached + 20K fresh input + 10K output tokens

Qwen3.5 397B
$0.168
Does not fit in one request
Cached input priced at the published list-input rate
Qwen3.6 Plus
API rate not published
Fits in one request
Cached-input rate unavailable

Qwen3.5 397B does not fit this workload in one request. Qwen3.5 397B has no published cached-input rate, so cached tokens use its listed input rate. Qwen3.6 Plus has no comparable published API token rate.

Specification differences

Sourced differences are shown directly. Missing facts stay explicit instead of being inferred from a model name or family.

Context window

Maximum documented context; output-token limits may be lower.

Qwen3.5 397B

128K

Qwen3.6 Plus

1M

API model ID

Qwen3.5 397B

Not sourced

Qwen3.6 Plus

Not sourced

Cached-input rate

A missing cached-input rate falls back to the listed input rate only in the stated workload estimate.

Qwen3.5 397B

Not published

Qwen3.6 Plus

No comparable hosted API rate

Documented inputs

Qwen3.5 397B

Not sourced

Qwen3.6 Plus

Not sourced

Documented outputs

Qwen3.5 397B

Not sourced

Qwen3.6 Plus

Not sourced

Provider availability

Qwen3.5 397B

Not sourced

Qwen3.6 Plus

Not sourced

Reasoning profile

Qwen3.5 397B

Non-Reasoning

Qwen3.6 Plus

Reasoning

Weight access

Qwen3.5 397B

Open Weight

Qwen3.6 Plus

Proprietary

License

Qwen3.5 397B

Open Weight

Qwen3.6 Plus

Proprietary

Release date

Qwen3.5 397B

2026-02-16

Qwen3.6 Plus

2026-04-02

If you already use one of these models
Deployment change
Both entries list Alibaba as the provider. Confirm endpoint, model ID, limits, and feature support before switching.
Quality signal
Qwen3.6 Plus has the higher public score estimate, 64.79 versus 57.46, but the 90% score intervals overlap.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Qwen3.6 Plus has the larger documented window (1M).

Run the same representative tasks against both endpoints before changing production traffic.

Benchmark evidence

The full public result ledger is available for audit without forcing a wide desktop table onto a phone.

Browse raw public benchmark evidence45 rows

Agentic

  • Terminal-Bench 2.0

    Shared source
    Qwen3.5 397B52.5%
    Qwen3.6 Plus61.6%

    Qwen3.6 Plus leads this result

  • BrowseComp

    Qwen3.5 397B62%
    Source
    Qwen3.6 Plus—

    Not directly comparable

  • Claw-Eval

    Shared source
    Qwen3.5 397B56.8%
    Qwen3.6 Plus58.8%

    Qwen3.6 Plus leads this result

  • QwenClawBench

    Shared source
    Qwen3.5 397B51.8%
    Qwen3.6 Plus57.2%

    Qwen3.6 Plus leads this result

  • τ³-bench results

    Shared source
    Qwen3.5 397B68.4%
    Qwen3.6 Plus70.7%

    Qwen3.6 Plus leads this result

  • VITA-Bench

    Shared source
    Qwen3.5 397B43.7%
    Qwen3.6 Plus44.3%

    Qwen3.6 Plus leads this result

  • DeepPlanning

    Shared source
    Qwen3.5 397B37.6%
    Qwen3.6 Plus41.5%

    Qwen3.6 Plus leads this result

  • Toolathlon

    Shared source
    Qwen3.5 397B36.3%
    Qwen3.6 Plus39.8%

    Qwen3.6 Plus leads this result

  • MCP Atlas

    Shared source
    Qwen3.5 397B46.1%
    Qwen3.6 Plus48.2%

    Qwen3.6 Plus leads this result

  • MCP-Tasks

    Shared source
    Qwen3.5 397B74.2%
    Qwen3.6 Plus74.1%

    Qwen3.5 397B leads this result

  • WideResearch

    Shared source
    Qwen3.5 397B74.0%
    Qwen3.6 Plus74.3%

    Qwen3.6 Plus leads this result

  • Gert Labs

    Shared source
    Qwen3.5 397B46.76%
    Qwen3.6 Plus50.60%

    Qwen3.6 Plus leads this result

  • ResearchClawBench

    Shared source
    Qwen3.5 397B14.2%
    Qwen3.6 Plus18.0%

    Qwen3.6 Plus leads this result

Coding

  • SWE-bench Verified

    Shared source
    Qwen3.5 397B76.2%
    Qwen3.6 Plus78.8%

    Qwen3.6 Plus leads this result

  • LiveCodeBench v6

    Shared source
    Qwen3.5 397B83.6%
    Qwen3.6 Plus87.1%

    Qwen3.6 Plus leads this result

  • SWE-bench Pro

    Shared source
    Qwen3.5 397B50.9%
    Qwen3.6 Plus56.6%

    Qwen3.6 Plus leads this result

  • SWE Multilingual

    Qwen3.5 397B—
    Qwen3.6 Plus73.8%
    Source

    Not directly comparable

  • Vibe Code Bench

    Qwen3.5 397B—
    Qwen3.6 Plus25.56%
    Source

    Not directly comparable

  • EEBench

    Qwen3.5 397B—
    Qwen3.6 Plus12.7%
    Source

    Not directly comparable

Reasoning

  • LongBench v2

    Shared source
    Qwen3.5 397B63.2%
    Qwen3.6 Plus62%

    Qwen3.5 397B leads this result

  • AI-Needle

    Shared source
    Qwen3.5 397B68.7%
    Qwen3.6 Plus68.3%

    Qwen3.5 397B leads this result

Knowledge

  • GPQA

    Shared source
    Qwen3.5 397B88.4%
    Qwen3.6 Plus90.4%

    Qwen3.6 Plus leads this result

  • SuperGPQA

    Shared source
    Qwen3.5 397B70.4%
    Qwen3.6 Plus71.6%

    Qwen3.6 Plus leads this result

  • MMLU-Pro

    Shared source
    Qwen3.5 397B87.8%
    Qwen3.6 Plus88.5%

    Qwen3.6 Plus leads this result

  • MMLU-Redux

    Shared source
    Qwen3.5 397B94.9%
    Qwen3.6 Plus94.5%

    Qwen3.5 397B leads this result

  • C-Eval

    Shared source
    Qwen3.5 397B93%
    Qwen3.6 Plus93.3%

    Qwen3.6 Plus leads this result

  • HLE

    Shared source
    Qwen3.5 397B28.7%
    Qwen3.6 Plus28.8%

    Qwen3.6 Plus leads this result

Math

  • AIME26

    Shared source
    Qwen3.5 397B93.3%
    Qwen3.6 Plus95.3%

    Qwen3.6 Plus leads this result

  • HMMT Feb 2025

    Shared source
    Qwen3.5 397B94.8%
    Qwen3.6 Plus96.7%

    Qwen3.6 Plus leads this result

  • HMMT Nov 2025

    Shared source
    Qwen3.5 397B92.7%
    Qwen3.6 Plus94.6%

    Qwen3.6 Plus leads this result

  • HMMT Feb 2026

    Shared source
    Qwen3.5 397B87.9%
    Qwen3.6 Plus87.8%

    Qwen3.5 397B leads this result

  • MMAnswerBench

    Shared source
    Qwen3.5 397B80.9%
    Qwen3.6 Plus83.8%

    Qwen3.6 Plus leads this result

  • FrontierMath v2 (Tiers 1-3)

    Qwen3.5 397B—
    Qwen3.6 Plus26.207%
    Source

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Qwen3.5 397B—
    Qwen3.6 Plus8.333%
    Source

    Not directly comparable

Multilingual

  • MMLU-ProX

    Shared source
    Qwen3.5 397B84.7%
    Qwen3.6 Plus84.7%

    Tie

  • NOVA-63

    Shared source
    Qwen3.5 397B59.1%
    Qwen3.6 Plus57.9%

    Qwen3.5 397B leads this result

Multimodal

  • MMMU-Pro

    Shared source
    Qwen3.5 397B79%
    Qwen3.6 Plus78.8%

    Qwen3.5 397B leads this result

  • MathVision

    Shared source
    Qwen3.5 397B88.6%
    Qwen3.6 Plus88.0%

    Qwen3.5 397B leads this result

  • CharXiv

    Shared source
    Qwen3.5 397B80.8%
    Qwen3.6 Plus81.5%

    Qwen3.6 Plus leads this result

  • VideoMMMU

    Shared source
    Qwen3.5 397B84.7%
    Qwen3.6 Plus84.0%

    Qwen3.5 397B leads this result

  • ScreenSpot Pro

    Shared source
    Qwen3.5 397B65.6%
    Qwen3.6 Plus68.2%

    Qwen3.6 Plus leads this result

  • V*

    Shared source
    Qwen3.5 397B95.8%
    Qwen3.6 Plus96.9%

    Qwen3.6 Plus leads this result

  • MMMU

    Qwen3.5 397B—
    Qwen3.6 Plus86.0%
    Source

    Not directly comparable

Instruction following

  • IFEval

    Shared source
    Qwen3.5 397B92.6%
    Qwen3.6 Plus94.3%

    Qwen3.6 Plus leads this result

  • IFBench

    Qwen3.5 397B—
    Qwen3.6 Plus75.8%
    Source

    Not directly comparable

Frequently asked questions

Which is better, Qwen3.5 397B or Qwen3.6 Plus?

Qwen3.6 Plus has the higher public score estimate, 64.79 versus 57.46, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Qwen3.5 397B or Qwen3.6 Plus?

Qwen3.6 Plus leads the like-for-like coding comparison across 2 shared weighted benchmark rows.

Which is better for agentic tasks, Qwen3.5 397B or Qwen3.6 Plus?

The current agentic tasks averages use different weighted benchmark sets, so BenchLM does not name a winner from them. Read the shared benchmark rows directly and test the models on the same task set.

Which costs less, Qwen3.5 397B or Qwen3.6 Plus?

Both models do not have comparable published API token rates, so this page does not name a universal price winner.

Which has the larger context window, Qwen3.5 397B or Qwen3.6 Plus?

Qwen3.6 Plus has the larger documented context window: 1M, compared with 128K.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated August 15, 2026

Watch Qwen3.5 397B vs Qwen3.6 Plus

One weekly email when material rank, price, or benchmark evidence changes make this matchup worth revisiting.

Read a sample issue

Join 2,000+ readers.