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
Qwen3.6 Plus

Alibaba

64.8/100

Supported · Public rank #36

90% interval 55.7–73.9

Qwen3.6 Plus vs Qwen3.7 Max

Updated August 18, 2026. Public scores include evidence status and uncertainty. They are not guarantees for a specific workload.

Model B
Qwen3.7 Max

Alibaba

71.6/100

Supported · Public rank #16

90% interval 64.6–78.6

Decision reading

Qwen3.7 Max has the higher public score estimate, 71.6 versus 64.79, but the 90% score intervals overlap. Treat that as a lead, not a settled winner.

20 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.

  • Agentic work

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

    Qwen3.7 Max

    Qwen3.7 Max leads on the same 1 weighted benchmark row.

    Confidence: limited

Show secondary and unsupported calls

  • Coding work

    Code generation, repair, and software-engineering tasks

    Not enough matched evidence

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

    Confidence: limited

  • Long documents

    Prompts that approach the documented context limit

    No clear pick

    The documented context windows are equal.

    Confidence: documented

  • 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

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

    Confidence: listed-rates

  • 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
20
Qwen3.6 Plus only
24
Qwen3.7 Max only
15
Like-for-like categories
4 / 8

2 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.

Agentic

Like-for-like
Qwen3.6 Plus
61.6
Qwen3.7 Max
69.7
Weighted basis
1 vs 1 rows
Reading
Qwen3.7 Max leads

Knowledge

Like-for-like
Qwen3.6 Plus
57.1
Qwen3.7 Max
64.2
Weighted basis
4 vs 4 rows
Reading
Qwen3.7 Max leads

Multilingual

Like-for-like
Qwen3.6 Plus
84.7
Qwen3.7 Max
87.0
Weighted basis
1 vs 1 rows
Reading
Qwen3.7 Max leads

Instruction following

Like-for-like
Qwen3.6 Plus
82.3
Qwen3.7 Max
84.4
Weighted basis
2 vs 2 rows
Reading
Qwen3.7 Max leads

Coding

Directional only
Qwen3.6 Plus
70.3
Qwen3.7 Max
77.9
Weighted basis
2 vs 4 rows
Reading
Directional only

Math

Directional only
Qwen3.6 Plus
60.5
Qwen3.7 Max
97.1
Weighted basis
4 vs 1 rows
Reading
Directional only

Reasoning

Not comparable
Qwen3.6 Plus
62.0
Qwen3.7 Max
90.4
Weighted basis
1 vs 1 rows
Reading
Not comparable

Multimodal

Not comparable
Qwen3.6 Plus
79.8
Qwen3.7 Max
Not measured
Weighted basis
2 vs 0 rows
Reading
Not comparable

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.

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

Qwen3.6 Plus
API rate not published
Fits in one request
Qwen3.7 Max
API rate not published
Fits in one request

Qwen3.6 Plus has no comparable published API token rate. Qwen3.7 Max has no comparable published API token rate.

Repository review

50K fresh input + 3K output tokens

Qwen3.6 Plus
API rate not published
Fits in one request
Qwen3.7 Max
API rate not published
Fits in one request

Qwen3.6 Plus has no comparable published API token rate. Qwen3.7 Max has no comparable published API token rate.

Cache-heavy agent loop

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

Qwen3.6 Plus
API rate not published
Fits in one request
Cached-input rate unavailable
Qwen3.7 Max
API rate not published
Fits in one request
Cached-input rate unavailable

Qwen3.6 Plus has no comparable published API token rate. Qwen3.7 Max 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.

Qwen3.6 Plus

1M

Qwen3.7 Max

1M

API model ID

Qwen3.6 Plus

Not sourced

Qwen3.7 Max

Not sourced

Cached-input rate

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

Qwen3.6 Plus

No comparable hosted API rate

Qwen3.7 Max

No comparable hosted API rate

Documented inputs

Qwen3.6 Plus

Not sourced

Qwen3.7 Max

Not sourced

Documented outputs

Qwen3.6 Plus

Not sourced

Qwen3.7 Max

Not sourced

Provider availability

Qwen3.6 Plus

Not sourced

Qwen3.7 Max

Not sourced

Reasoning profile

Qwen3.6 Plus

Reasoning

Qwen3.7 Max

Reasoning

Weight access

Qwen3.6 Plus

Proprietary

Qwen3.7 Max

Proprietary

License

Qwen3.6 Plus

Proprietary

Qwen3.7 Max

Proprietary

Release date

Qwen3.6 Plus

2026-04-02

Qwen3.7 Max

2026-05-16

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
Qwen3.7 Max has the higher public score estimate, 71.6 versus 64.79, but the 90% score intervals overlap.
Workload cost
A complete comparable API-rate estimate is not available for both models.
Context tradeoff
Both models list 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 evidence59 rows

Agentic

  • Terminal-Bench 2.0

    Qwen3.6 Plus61.6%
    Source
    Qwen3.7 Max69.7%
    Source

    Qwen3.7 Max leads this result

  • Claw-Eval

    Qwen3.6 Plus58.8%
    Source
    Qwen3.7 Max65.2%
    Source

    Qwen3.7 Max leads this result

  • QwenClawBench

    Qwen3.6 Plus57.2%
    Source
    Qwen3.7 Max64.3%
    Source

    Qwen3.7 Max leads this result

  • τ³-bench results

    Qwen3.6 Plus70.7%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • VITA-Bench

    Qwen3.6 Plus44.3%
    Source
    Qwen3.7 Max47.9%
    Source

    Qwen3.7 Max leads this result

  • DeepPlanning

    Qwen3.6 Plus41.5%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • Toolathlon

    Qwen3.6 Plus39.8%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • MCP Atlas

    Qwen3.6 Plus48.2%
    Source
    Qwen3.7 Max76.4%
    Source

    Qwen3.7 Max leads this result

  • MCP-Tasks

    Qwen3.6 Plus74.1%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • WideResearch

    Qwen3.6 Plus74.3%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • Gert Labs

    Shared source
    Qwen3.6 Plus50.60%
    Qwen3.7 Max64.27%

    Qwen3.7 Max leads this result

  • ResearchClawBench

    Shared source
    Qwen3.6 Plus18.0%
    Qwen3.7 Max18.7%

    Qwen3.7 Max leads this result

  • BFCL v4

    Qwen3.6 Plus—
    Qwen3.7 Max75.0%
    Source

    Not directly comparable

  • HLE w/ tools

    Qwen3.6 Plus—
    Qwen3.7 Max53.5%
    Source

    Not directly comparable

Coding

  • SWE-bench Verified

    Qwen3.6 Plus78.8%
    Source
    Qwen3.7 Max80.4%
    Source

    Qwen3.7 Max leads this result

  • SWE-bench Pro

    Qwen3.6 Plus56.6%
    Source
    Qwen3.7 Max60.6%
    Source

    Qwen3.7 Max leads this result

  • SWE Multilingual

    Qwen3.6 Plus73.8%
    Source
    Qwen3.7 Max78.3%
    Source

    Qwen3.7 Max leads this result

  • LiveCodeBench v6

    Qwen3.6 Plus87.1%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • Vibe Code Bench

    Qwen3.6 Plus25.56%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • EEBench

    Qwen3.6 Plus12.7%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • NL2Repo

    Qwen3.6 Plus—
    Qwen3.7 Max47.2%
    Source

    Not directly comparable

  • SciCode

    Qwen3.6 Plus—
    Qwen3.7 Max53.5%
    Source

    Not directly comparable

  • LiveCodeBench

    Qwen3.6 Plus—
    Qwen3.7 Max91.6%
    Source

    Not directly comparable

  • Terminal-Bench 2.0

    Qwen3.6 Plus—
    Qwen3.7 Max69.7%
    Source

    Not directly comparable

Reasoning

  • AI-Needle

    Qwen3.6 Plus68.3%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • LongBench v2

    Qwen3.6 Plus62%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • MRCRv2

    Qwen3.6 Plus—
    Qwen3.7 Max90.4%
    Source

    Not directly comparable

  • CritPt

    Qwen3.6 Plus—
    Qwen3.7 Max13.4%
    Source

    Not directly comparable

Knowledge

  • GPQA

    Qwen3.6 Plus90.4%
    Source
    Qwen3.7 Max92.4%
    Source

    Qwen3.7 Max leads this result

  • SuperGPQA

    Qwen3.6 Plus71.6%
    Source
    Qwen3.7 Max73.6%
    Source

    Qwen3.7 Max leads this result

  • MMLU-Pro

    Qwen3.6 Plus88.5%
    Source
    Qwen3.7 Max89.6%
    Source

    Qwen3.7 Max leads this result

  • MMLU-Redux

    Qwen3.6 Plus94.5%
    Source
    Qwen3.7 Max95%
    Source

    Qwen3.7 Max leads this result

  • C-Eval

    Qwen3.6 Plus93.3%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • HLE

    Qwen3.6 Plus28.8%
    Source
    Qwen3.7 Max41.4%
    Source

    Qwen3.7 Max leads this result

  • GPQA-D

    Qwen3.6 Plus—
    Qwen3.7 Max92.4%
    Source

    Not directly comparable

  • MMMLU

    Qwen3.6 Plus—
    Qwen3.7 Max90.3%
    Source

    Not directly comparable

Math

  • AIME26

    Qwen3.6 Plus95.3%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • HMMT Feb 2025

    Qwen3.6 Plus96.7%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • HMMT Nov 2025

    Qwen3.6 Plus94.6%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • HMMT Feb 2026

    Qwen3.6 Plus87.8%
    Source
    Qwen3.7 Max97.1%
    Source

    Qwen3.7 Max leads this result

  • MMAnswerBench

    Qwen3.6 Plus83.8%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • FrontierMath v2 (Tiers 1-3)

    Qwen3.6 Plus26.207%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • FrontierMath v2 (Tier 4)

    Qwen3.6 Plus8.333%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • IMOAnswerBench

    Qwen3.6 Plus—
    Qwen3.7 Max90.0%
    Source

    Not directly comparable

  • Apex

    Qwen3.6 Plus—
    Qwen3.7 Max44.5%
    Source

    Not directly comparable

Multilingual

  • MMLU-ProX

    Qwen3.6 Plus84.7%
    Source
    Qwen3.7 Max87%
    Source

    Qwen3.7 Max leads this result

  • NOVA-63

    Qwen3.6 Plus57.9%
    Source
    Qwen3.7 Max59.0%
    Source

    Qwen3.7 Max leads this result

  • INCLUDE

    Qwen3.6 Plus—
    Qwen3.7 Max86.2%
    Source

    Not directly comparable

  • MAXIFE

    Qwen3.6 Plus—
    Qwen3.7 Max89.2%
    Source

    Not directly comparable

  • PolyMath

    Qwen3.6 Plus—
    Qwen3.7 Max86.5%
    Source

    Not directly comparable

Multimodal

  • MMMU

    Qwen3.6 Plus86.0%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • MMMU-Pro

    Qwen3.6 Plus78.8%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • MathVision

    Qwen3.6 Plus88.0%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • VideoMMMU

    Qwen3.6 Plus84.0%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • ScreenSpot Pro

    Qwen3.6 Plus68.2%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • CharXiv

    Qwen3.6 Plus81.5%
    Source
    Qwen3.7 Max—

    Not directly comparable

  • V*

    Qwen3.6 Plus96.9%
    Source
    Qwen3.7 Max—

    Not directly comparable

Instruction following

Frequently asked questions

Which is better, Qwen3.6 Plus or Qwen3.7 Max?

Qwen3.7 Max has the higher public score estimate, 71.6 versus 64.79, but the 90% score intervals overlap. The higher estimate is not a decisive winner because the uncertainty ranges overlap.

Which is better for coding, Qwen3.6 Plus or Qwen3.7 Max?

The current coding 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 is better for agentic tasks, Qwen3.6 Plus or Qwen3.7 Max?

Qwen3.7 Max leads the like-for-like agentic tasks comparison across 1 shared weighted benchmark row.

Which costs less, Qwen3.6 Plus or Qwen3.7 Max?

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, Qwen3.6 Plus or Qwen3.7 Max?

Both models list the same context window, 1M.

Related comparisons

Compare API pricing Read the methodology Open the model selector

Last updated August 18, 2026

Watch Qwen3.6 Plus vs Qwen3.7 Max

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

Read a sample issue

Join 2,000+ readers.