Skip to main content
BenchLM
Advertise

Model comparison

Qwen3.5-27B vs Qwen3.6-27B

Data verified

Head-to-head evidence from 20 shared benchmark results across 6 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

60.7/100
Margin
6.9pts
← winning
53.82/100
1 category wins2 category wins

Public leaderboard positions: Qwen3.5-27B #45 (Supported); Qwen3.6-27B #93 (Estimated). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Qwen3.5-27B and Qwen3.6-27B share 20 comparable benchmark results. 3 of 8 categories are comparable. 8 results are unique to Qwen3.5-27B; 34 to Qwen3.6-27B.

Updated July 21, 2026
Shared results
20
Qwen3.5-27B only
8
Qwen3.6-27B only
34
Comparable categories
3 / 8

Pick Qwen3.5-27B if you want the stronger benchmark profile. Qwen3.6-27B only becomes the better choice if coding is the priority.

Confidence note. This is a partial-evidence comparison with 20 shared benchmark results across 6 evidence categories; 3 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.

Why this result

Qwen3.5-27B is clearly ahead on the BenchAlign aggregate, 60.7 to 53.82. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.

Qwen3.5-27B's sharpest advantage is in knowledge, where it averages 82.7 against 53.3. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 41.6% to 59.3%. Qwen3.6-27B does hit back in coding, so the answer changes if that is the part of the workload you care about most.

Category breakdown

Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.

Category scores and score margins for Qwen3.5-27B and Qwen3.6-27B
CategoryQwen3.5-27BΔQwen3.6-27B
KnowledgeQwen3.5-27B82.7Margin← 29.4Qwen3.6-27B53.3
CodingQwen3.5-27B64.9Margin→ 12.6Qwen3.6-27B77.5
AgenticQwen3.5-27B52.0Margin→ 7.3Qwen3.6-27B59.3
ReasoningQwen3.5-27B60.6MarginNo overlapQwen3.6-27BNot measured
MathQwen3.5-27BNot measuredMarginNo overlapQwen3.6-27B89.2
MultilingualQwen3.5-27B82.2MarginNo overlapQwen3.6-27BNot measured
MultimodalQwen3.5-27BNot measuredMarginNo overlapQwen3.6-27B76.7
Inst. FollowingQwen3.5-27B95.0MarginNo overlapQwen3.6-27BNot measured

Decisive benchmark drivers

The largest measured benchmark gaps in this matchup, with exact reported values.

More
A · Qwen3.5-27BB · Qwen3.6-27B

  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 41.6%B 59.3%
    Winner: Qwen3.6-27BΔ 17.7
    Terminal-Bench 2.0: Qwen3.5-27B scored 41.6%; Qwen3.6-27B scored 59.3%. Qwen3.6-27B wins this benchmark.

  2. SWE-bench Verified

    Coding
    Source ↗
    A 72.4%B 77.2%
    Winner: Qwen3.6-27BΔ 4.8
    SWE-bench Verified: Qwen3.5-27B scored 72.4%; Qwen3.6-27B scored 77.2%. Qwen3.6-27B wins this benchmark.

  3. GPQA

    Knowledge
    Source ↗
    A 85.5%B 87.8%
    Winner: Qwen3.6-27BΔ 2.3
    GPQA: Qwen3.5-27B scored 85.5%; Qwen3.6-27B scored 87.8%. Qwen3.6-27B wins this benchmark.

  4. SuperGPQA

    Knowledge
    Source ↗
    A 65.6%B 66%
    Winner: Qwen3.6-27BΔ 0.4
    SuperGPQA: Qwen3.5-27B scored 65.6%; Qwen3.6-27B scored 66%. Qwen3.6-27B wins this benchmark.

  5. MMLU-Pro

    Knowledge
    Source ↗
    A 86.1%B 86.2%
    Winner: Qwen3.6-27BΔ 0.1
    MMLU-Pro: Qwen3.5-27B scored 86.1%; Qwen3.6-27B scored 86.2%. Qwen3.6-27B wins this benchmark.

Operational comparison

Runtime and commercial metrics are compared only when both models have a complete sourced value.

MetricQwen3.5-27BQwen3.6-27BComparison
Input / output priceUSD per 1M tokensQwen3.5-27B$0 input / $0 outputQwen3.6-27B$0 input / $0 outputListed prices are equal.
Generation speedtokens per secondQwen3.5-27BNot availableQwen3.6-27BNot availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenQwen3.5-27BNot availableQwen3.6-27BNot availableA complete latency comparison is not available.
Context windowmaximum listed tokensQwen3.5-27B262KQwen3.6-27B262KListed context windows are equal.

Benchmark Deep Dive

AgenticQwen3.6-27B wins
BenchmarkQwen3.5-27BQwen3.6-27BResult
Terminal-Bench 2.0Source 41.6%59.3%Qwen3.6-27B leads
BrowseCompSource 61%—Not comparable
OSWorld-VerifiedSource 56.2%—Not comparable
τ²-bench resultsSource 93.9%94.2%Qwen3.6-27B leads
Gert LabsSource 39.41%54.84%Qwen3.6-27B leads
Claw-EvalSource —72.4%Not comparable
QwenClawBenchSource —53.4%Not comparable
QwenWebBenchSource —1487Not comparable
AndroidWorldSource —70.3%Not comparable
AA Agentic IndexSource —27.0%Not comparable
GDPval-AASource —32.0%Not comparable
GDPval-AASource —1140Not comparable
CodingQwen3.6-27B wins
BenchmarkQwen3.5-27BQwen3.6-27BResult
SWE-bench VerifiedSource 72.4%77.2%Qwen3.6-27B leads
SWE-RebenchSource 58.9%—Not comparable
AA-SciCodeSource 39.5%39.8%Qwen3.6-27B leads
SWE MultilingualSource —71.3%Not comparable
SWE-bench ProSource —53.5%Not comparable
Terminal-Bench 2.0Source —59.3%Not comparable
LiveCodeBenchSource —83.9%Not comparable
NL2RepoSource —36.2%Not comparable
AA Coding IndexSource —53.7%Not comparable
Reasoning
BenchmarkQwen3.5-27BQwen3.6-27BResult
LongBench v2Source 60.6%—Not comparable
AA-LCRSource 67.3%68.7%Qwen3.6-27B leads
CritPtSource 0.9%1.1%Qwen3.6-27B leads
KnowledgeQwen3.5-27B wins
BenchmarkQwen3.5-27BQwen3.6-27BResult
MMLU-ProSource 86.1%86.2%Qwen3.6-27B leads
SuperGPQASource 65.6%66%Qwen3.6-27B leads
GPQASource 85.5%87.8%Qwen3.6-27B leads
Artificial Analysis Intelligence IndexSource 33.8%37.0%Qwen3.6-27B leads
AA-GPQA DiamondSource 85.8%84.2%Qwen3.5-27B leads
AA-HLESource 22.2%21.6%Qwen3.5-27B leads
AA-Omniscience IndexSource -42.0%-19.8%Qwen3.6-27B leads
AA-Omniscience AccuracySource 21.0%19.2%Qwen3.5-27B leads
AA-Omniscience Hallucination RateSource 79.7%48.3%Qwen3.6-27B leads
MMLU-ReduxSource —93.5%Not comparable
C-EvalSource —91.4%Not comparable
HLESource —24%Not comparable
Math
BenchmarkQwen3.5-27BQwen3.6-27BResult
HMMT Feb 2025Source —93.8%Not comparable
HMMT Nov 2025Source —90.7%Not comparable
HMMT Feb 2026Source —84.3%Not comparable
MMAnswerBenchSource —80.8%Not comparable
AIME26Source —94.1%Not comparable
Multilingual
BenchmarkQwen3.5-27BQwen3.6-27BResult
MMLU-ProXSource 82.2%—Not comparable
Multimodal
BenchmarkQwen3.5-27BQwen3.6-27BResult
MMMUSource 82.3%82.9%Qwen3.6-27B leads
MMVUSource 73.3%—Not comparable
MathVisionSource 86.0%—Not comparable
V*Source 93.7%94.7%Qwen3.6-27B leads
AA-MMMU-ProSource 75.0%74.6%Qwen3.5-27B leads
MMMU-ProSource —75.8%Not comparable
RealWorldQASource —84.1%Not comparable
DynaMathSource —85.6%Not comparable
MStarSource —81.4%Not comparable
SimpleVQASource —56.1%Not comparable
CharXivSource —78.4%Not comparable
CC-OCRSource —81.2%Not comparable
CountBenchSource —97.8%Not comparable
RefCOCO (avg)Source —92.5%Not comparable
ERQASource —62.5%Not comparable
Video-MME (with subtitle)Source —87.7%Not comparable
VideoMMMUSource —84.4%Not comparable
MLVU (M-Avg)Source —86.6%Not comparable
Inst. Following
BenchmarkQwen3.5-27BQwen3.6-27BResult
IFEvalSource 95%—Not comparable
AA-IFBenchSource 75.6%67.6%Qwen3.5-27B leads
Frequently Asked Questions (4)

Which is better, Qwen3.5-27B or Qwen3.6-27B?

Qwen3.5-27B is ahead on BenchLM's BenchAlign leaderboard, 60.7 to 53.82. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 41.6% and 59.3%.

Which is better for knowledge tasks, Qwen3.5-27B or Qwen3.6-27B?

Qwen3.5-27B has the edge for knowledge tasks in this comparison, averaging 82.7 versus 53.3. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.

Which is better for coding, Qwen3.5-27B or Qwen3.6-27B?

Qwen3.6-27B has the edge for coding in this comparison, averaging 77.5 versus 64.9. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Qwen3.5-27B or Qwen3.6-27B?

Qwen3.6-27B has the edge for agentic tasks in this comparison, averaging 59.3 versus 52. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.

Self-host vs API cost

Estimates at 50,000 req/day · 1000 tokens/req average.

Qwen3.5-27B
API / mo$0
Self-host / moNot listed
Break-even—
Proprietary model — self-hosting not applicable.
Qwen3.6-27B
API / mo$0
Self-host / mo$429
Break-even—
Model the full break-even

Related Comparisons

Explore More

Alibaba Compare Pricing Methodology Find Your Best LLM Overall Rankings
Last updated: July 21, 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.