Skip to main content

Model comparison

Qwen3.5-122B-A10B vs Qwen3.5-27B

Data verified

Head-to-head evidence from 26 shared benchmark results across 7 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.

60.56/100
Margin
0.1pts
winning →
60.7/100
3 category wins2 category wins

Public leaderboard positions: Qwen3.5-122B-A10B #47 (Supported); Qwen3.5-27B #45 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.

Evidence parity. Qwen3.5-122B-A10B and Qwen3.5-27B share 26 comparable benchmark results. 6 of 8 categories are comparable. 5 results are unique to Qwen3.5-122B-A10B; 2 to Qwen3.5-27B.

Updated July 23, 2026
Shared results
26
Qwen3.5-122B-A10B only
5
Qwen3.5-27B only
2
Comparable categories
6 / 8

Pick Qwen3.5-27B if you want the stronger benchmark profile. Qwen3.5-122B-A10B only becomes the better choice if coding is the priority.

Confidence note. This is a partial-evidence comparison with 26 shared benchmark results across 7 evidence categories; 6 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 has the cleaner BenchAlign overall profile here, landing at 60.7 versus 60.56. It is a real lead, but still close enough that category-level strengths matter more than the headline number.

Qwen3.5-27B's sharpest advantage is in instruction following, where it averages 95 against 93.4. The single biggest benchmark swing on the page is Terminal-Bench 2.0, 49.4% to 41.6%. Qwen3.5-122B-A10B 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 Qwen3.5-122B-A10B and Qwen3.5-27B
CategoryQwen3.5-122B-A10BΔQwen3.5-27B
CodingQwen3.5-122B-A10B72.0Margin 7.1Qwen3.5-27B64.9
AgenticQwen3.5-122B-A10B56.4Margin 4.4Qwen3.5-27B52.0
Inst. FollowingQwen3.5-122B-A10B93.4Margin 1.6Qwen3.5-27B95.0
KnowledgeQwen3.5-122B-A10B83.6Margin 0.9Qwen3.5-27B82.7
ReasoningQwen3.5-122B-A10B60.2Margin 0.4Qwen3.5-27B60.6
MultilingualQwen3.5-122B-A10B82.2MarginTieQwen3.5-27B82.2
MultimodalQwen3.5-122B-A10B77.2MarginNo overlapQwen3.5-27BNot measured

Decisive benchmark drivers

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

More
A · Qwen3.5-122B-A10BB · Qwen3.5-27B
  1. Terminal-Bench 2.0

    Agentic
    Source ↗
    A 49.4%B 41.6%
    Winner: Qwen3.5-122B-A10BΔ 7.8
    Terminal-Bench 2.0: Qwen3.5-122B-A10B scored 49.4%; Qwen3.5-27B scored 41.6%. Qwen3.5-122B-A10B wins this benchmark.
  2. BrowseComp

    Agentic
    Source ↗
    A 63.8%B 61%
    Winner: Qwen3.5-122B-A10BΔ 2.8
    BrowseComp: Qwen3.5-122B-A10B scored 63.8%; Qwen3.5-27B scored 61%. Qwen3.5-122B-A10B wins this benchmark.
  3. OSWorld-Verified

    Agentic
    Source ↗
    A 58%B 56.2%
    Winner: Qwen3.5-122B-A10BΔ 1.8
    OSWorld-Verified: Qwen3.5-122B-A10B scored 58%; Qwen3.5-27B scored 56.2%. Qwen3.5-122B-A10B wins this benchmark.
  4. IFEval

    Inst. Following
    Source ↗
    A 93.4%B 95%
    Winner: Qwen3.5-27BΔ 1.6
    IFEval: Qwen3.5-122B-A10B scored 93.4%; Qwen3.5-27B scored 95%. Qwen3.5-27B wins this benchmark.
  5. SuperGPQA

    Knowledge
    Source ↗
    A 67.1%B 65.6%
    Winner: Qwen3.5-122B-A10BΔ 1.5
    SuperGPQA: Qwen3.5-122B-A10B scored 67.1%; Qwen3.5-27B scored 65.6%. Qwen3.5-122B-A10B wins this benchmark.

Operational comparison

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

MetricQwen3.5-122B-A10BQwen3.5-27BComparison
Input / output priceUSD per 1M tokensQwen3.5-122B-A10B$0 input / $0 outputQwen3.5-27B$0 input / $0 outputListed prices are equal.
Generation speedtokens per secondQwen3.5-122B-A10BNot availableQwen3.5-27BNot availableA complete speed comparison is not available.
First-answer latencyseconds to first tokenQwen3.5-122B-A10BNot availableQwen3.5-27BNot availableA complete latency comparison is not available.
Context windowmaximum listed tokensQwen3.5-122B-A10B262KQwen3.5-27B262KListed context windows are equal.

Benchmark Deep Dive

AgenticQwen3.5-122B-A10B wins
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
Terminal-Bench 2.0Source 49.4%41.6%Qwen3.5-122B-A10B leads
BrowseCompSource 63.8%61%Qwen3.5-122B-A10B leads
OSWorld-VerifiedSource 58%56.2%Qwen3.5-122B-A10B leads
τ²-bench resultsSource 93.6%93.9%Qwen3.5-27B leads
AA Agentic IndexSource 20.7%Not comparable
GDPval-AASource 23.9%Not comparable
GDPval-AASource 978Not comparable
Gert LabsSource 39.41%Not comparable
CodingQwen3.5-122B-A10B wins
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
SWE-bench VerifiedSource 72%72.4%Qwen3.5-27B leads
AA Coding IndexSource 45.7%Not comparable
AA-SciCodeSource 42.0%39.5%Qwen3.5-122B-A10B leads
SWE-RebenchSource 58.9%Not comparable
ReasoningQwen3.5-27B wins
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
LongBench v2Source 60.2%60.6%Qwen3.5-27B leads
AA-LCRSource 66.7%67.3%Qwen3.5-27B leads
CritPtSource 0.6%0.9%Qwen3.5-27B leads
KnowledgeQwen3.5-122B-A10B wins
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
MMLU-ProSource 86.7%86.1%Qwen3.5-122B-A10B leads
SuperGPQASource 67.1%65.6%Qwen3.5-122B-A10B leads
GPQASource 86.6%85.5%Qwen3.5-122B-A10B leads
Artificial Analysis Intelligence IndexSource 32.3%33.8%Qwen3.5-27B leads
AA-GPQA DiamondSource 85.7%85.8%Qwen3.5-27B leads
AA-HLESource 23.4%22.2%Qwen3.5-122B-A10B leads
AA-Omniscience IndexSource -39.6%-42.0%Qwen3.5-122B-A10B leads
AA-Omniscience AccuracySource 24.7%21.0%Qwen3.5-122B-A10B leads
AA-Omniscience Hallucination RateSource 85.5%79.7%Qwen3.5-27B leads
MultilingualTie
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
MMLU-ProXSource 82.2%82.2%Tie
Multimodal
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
MMMUSource 83.9%82.3%Qwen3.5-122B-A10B leads
MMVUSource 74.7%73.3%Qwen3.5-122B-A10B leads
MathVisionSource 86.2%86.0%Qwen3.5-122B-A10B leads
CharXivSource 77.2%Not comparable
V*Source 93.2%93.7%Qwen3.5-27B leads
AA-MMMU-ProSource 75.0%75.0%Tie
Inst. FollowingQwen3.5-27B wins
BenchmarkQwen3.5-122B-A10BQwen3.5-27BResult
IFEvalSource 93.4%95%Qwen3.5-27B leads
AA-IFBenchSource 75.7%75.6%Qwen3.5-122B-A10B leads
Frequently Asked Questions (7)

Which is better, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-27B is ahead on BenchLM's BenchAlign leaderboard, 60.7 to 60.56. The biggest single separator in this matchup is Terminal-Bench 2.0, where the scores are 49.4% and 41.6%.

Which is better for knowledge tasks, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-122B-A10B has the edge for knowledge tasks in this comparison, averaging 83.6 versus 82.7. Inside this category, AA-Omniscience Hallucination Rate is the benchmark that creates the most daylight between them.

Which is better for coding, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-122B-A10B has the edge for coding in this comparison, averaging 72 versus 64.9. Inside this category, AA-SciCode is the benchmark that creates the most daylight between them.

Which is better for reasoning, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-27B has the edge for reasoning in this comparison, averaging 60.6 versus 60.2. Inside this category, AA-LCR is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-122B-A10B has the edge for agentic tasks in this comparison, averaging 56.4 versus 52. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.

Which is better for instruction following, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-27B has the edge for instruction following in this comparison, averaging 95 versus 93.4. Inside this category, IFEval is the benchmark that creates the most daylight between them.

Which is better for multilingual tasks, Qwen3.5-122B-A10B or Qwen3.5-27B?

Qwen3.5-122B-A10B and Qwen3.5-27B are effectively tied for multilingual tasks here, both landing at 82.2 on average.

Related Comparisons

Last updated: July 23, 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.