LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

GPT-5.5 vs Qwen3.8 27B

Updated October 21, 2026. 82 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace GPT-5.5 Replace Qwen3.8 27B
GPT-5.5OpenAI
Coding
67.8
Overall
38.4
Output $/M
$30.00
72
Qwen3.8 27BQwen
Coding
73.6
Overall
33.7
Output $/M
$1.35
10
vs boards won of 82 shared

Decision reading

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 67.8 for GPT-5.5.

Pick GPT-5.5 when the context window is larger, 1.05M against 1M tokens, or when what matters is AA Coding Index or LiveCodeBench (Vals AI), where it leads.

Model AGPT-5.5OpenAI67.8/1002 independent benchmarks$1.65 per solved taskModel BQwen3.8 27BQwen73.6/1002 independent benchmarks$1.26 per solved task
Shared results
82
GPT-5.5 only
50
Qwen3.8 27B only
3
Like-for-like categories
8 of 9

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

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
GPT-5.5 67.8 from 2 benchmarks, Qwen3.8 27B 73.6 from 2 benchmarks.
Widest category
Math, 15.3 points apart on average over 3 shared benchmarks.
Widest single gap
Surface Evolver Bench: GPT-5.5 88.1% against Qwen3.8 27B 45.0%, 43.1 points apart on a 0 to 100 scale.
Output price
GPT-5.5 $30.00 against Qwen3.8 27B $1.35 per million output tokens, 22.2x apart.
Context window
GPT-5.5 takes 1.05M tokens, Qwen3.8 27B 1M.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryGPT-5.5BarsQwen3.8 27BTies
Coding106
Intelligence20
Agentic work72
Reasoning120
Math30
Knowledge90
Web design01
General291
All shared boards7210
CodingLike-for-likeFour of these boards feed the Vibe Coding Index106
Coding benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
LiveBench coding85.1%86.8%Qwen3.8 27B
LMArena WebDev15131593Qwen3.8 27B
AA Coding Index74.968.1GPT-5.5
LiveCodeBench (Vals AI)85.3%84.0%GPT-5.5
SWE-bench Verified (Vals AI)82.6%86.0%Qwen3.8 27B
Vibe Code Bench (Vals AI)69.8%64.8%GPT-5.5
Code migration (Vals AI)45.2%14.2%GPT-5.5
ProgramBench (Vals AI)0.5%0.0%GPT-5.5
IOI (Vals AI)Not measured39.1%Only Qwen3.8 27B measured
Terminal-Bench 2.1 (Vals AI)76.4%58.4%GPT-5.5
SkillsBench (Vals AI)62.5%38.1%GPT-5.5
CyberBench (Vals AI)Not measured67.8%Only Qwen3.8 27B measured
SciCode56.1%46.6%GPT-5.5
Terminal-Bench Hard (Artificial Analysis run)60.6%Not measuredOnly GPT-5.5 measured
Terminal-Bench84.7%Not measuredOnly GPT-5.5 measured
FrontierCode43.0%Not measuredOnly GPT-5.5 measured
GSO-Bench40.2%Not measuredOnly GPT-5.5 measured
WeirdML84.9%Not measuredOnly GPT-5.5 measured
MirrorCode10.0%Not measuredOnly GPT-5.5 measured
ALE-Bench1943.0Not measuredOnly GPT-5.5 measured
LMArena Text: coding14941483GPT-5.5
LMArena Text: software and IT14931478GPT-5.5
LMArena WebDev: React15061602Qwen3.8 27B
LMArena WebDev: HTML15431570Qwen3.8 27B
LMArena WebDev: from an image15321577Qwen3.8 27B
Neither model has a result on the other 13 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
AA Intelligence Index38.433.7GPT-5.5
Epoch Capabilities Index159.7151.7GPT-5.5
Agentic workLike-for-likePublic boards, shown as published, kept out of the index72
Agentic work benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
AA Agentic Index36.445.8Qwen3.8 27B
Design Arena: agents building web apps1117Not measuredOnly GPT-5.5 measured
Design Arena: agents building full-stack apps1089Not measuredOnly GPT-5.5 measured
Design Arena: agents building mobile apps1143Not measuredOnly GPT-5.5 measured
Design Arena: agents building Android apps1169Not measuredOnly GPT-5.5 measured
Design Arena: agents building Godot games1173Not measuredOnly GPT-5.5 measured
Design Arena: agents building games1179Not measuredOnly GPT-5.5 measured
Design Arena: HTML slides1067Not measuredOnly GPT-5.5 measured
Design Arena: PowerPoint slides1108Not measuredOnly GPT-5.5 measured
LiveBench research61.6%59.4%GPT-5.5
LiveBench browser48.0%32.0%GPT-5.5
LiveBench long agentic44.6%27.9%GPT-5.5
SREBench (Vals AI)3.8%Not measuredOnly GPT-5.5 measured
Finance Agent (Vals AI)51.8%48.5%GPT-5.5
Tax Agent Bench (Vals AI)21.8%30.9%Qwen3.8 27B
Legal Research Bench (Vals AI)40.4%36.1%GPT-5.5
Excel Modeling Benchmark (Vals AI)64.5%59.7%GPT-5.5
Terminal-Bench-Science (Vals AI)Not measured0.0%Only Qwen3.8 27B measured
Time Horizon Index (Vals AI)8.3%Not measuredOnly GPT-5.5 measured
τ²-Bench Telecom93.9%Not measuredOnly GPT-5.5 measured
APEX-Agents55.1%47.5%GPT-5.5
Remote Labor Index6.3%Not measuredOnly GPT-5.5 measured
OSWorld 2.013.0%Not measuredOnly GPT-5.5 measured
DeepResearch Bench54.0%Not measuredOnly GPT-5.5 measured
PostTrainBench27.2%Not measuredOnly GPT-5.5 measured
Vending-Bench 2$7,524Not measuredOnly GPT-5.5 measured
Neither model has a result on the other 13 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index120
Reasoning benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
GPQA Diamond93.2%88.9%GPT-5.5
LiveBench reasoning89.7%80.0%GPT-5.5
MMMU Pro (Vals AI)88.3%83.9%GPT-5.5
GPQA Diamond (Artificial Analysis run)93.5%90.5%GPT-5.5
Humanity's Last Exam (Artificial Analysis run)45.8%33.9%GPT-5.5
GPQA Diamond (Epoch AI run)90.7%Not measuredOnly GPT-5.5 measured
ARC-AGI95.0%87.5%GPT-5.5
ARC-AGI-285.0%42.4%GPT-5.5
SimpleBench69.0%Not measuredOnly GPT-5.5 measured
Chess puzzles63.0%Not measuredOnly GPT-5.5 measured
Mystery game puzzles56.0%Not measuredOnly GPT-5.5 measured
EBR-bench34.3%Not measuredOnly GPT-5.5 measured
LMArena Text: hard prompts14891459GPT-5.5
LMArena Text: expert15071479GPT-5.5
LMArena Text: hard prompts in English14831461GPT-5.5
LMArena Vision: diagrams13251285GPT-5.5
LMArena Vision: homework13321295GPT-5.5
Neither model has a result on the other 4 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index30
Math benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
OTIS Mock AIME84.4%Not measuredOnly GPT-5.5 measured
FrontierMath, tiers 1 to 385.3%Not measuredOnly GPT-5.5 measured
FrontierMath, tier 472.5%Not measuredOnly GPT-5.5 measured
ProofBench50.0%16.0%GPT-5.5
LMArena Text: math14871455GPT-5.5
LMArena Text: mathematical work14961458GPT-5.5
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index90
Knowledge benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
MMLU Pro88.1%84.3%GPT-5.5
LegalBench (Vals AI)86.5%82.4%GPT-5.5
CaseLaw (Vals AI)66.2%Not measuredOnly GPT-5.5 measured
CorpFin (Vals AI)68.4%Not measuredOnly GPT-5.5 measured
TaxEval (Vals AI)75.0%70.8%GPT-5.5
MortgageTax (Vals AI)68.8%64.9%GPT-5.5
MedCode (Vals AI)49.1%28.7%GPT-5.5
SimpleQA Verified63.0%Not measuredOnly GPT-5.5 measured
LMArena Text: science14851467GPT-5.5
LMArena Text: medicine and healthcare14651437GPT-5.5
LMArena Text: legal and government14911447GPT-5.5
LMArena Text: business and finance14781438GPT-5.5
Neither model has a result on the other 5 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index01
Web design benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
Design Arena: websites1256Not measuredOnly GPT-5.5 measured
Design Arena: UI components1260Not measuredOnly GPT-5.5 measured
Design Arena: game dev1294Not measuredOnly GPT-5.5 measured
Design Arena: data viz1265Not measuredOnly GPT-5.5 measured
Design Arena: 3D1198Not measuredOnly GPT-5.5 measured
Design Arena: all code categories1260Not measuredOnly GPT-5.5 measured
Design Arena: SVG1234Not measuredOnly GPT-5.5 measured
Design Arena: ASCII art1254Not measuredOnly GPT-5.5 measured
LiveBench design70.3%75.9%Qwen3.8 27B
GeneralLike-for-likePublic boards, shown as published, kept out of the index291
General benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
LiveBench data analysis81.6%76.6%GPT-5.5
Public Benefits Bench (Vals AI)60.9%Not measuredOnly GPT-5.5 measured
SAGE (Vals AI)51.5%52.4%Qwen3.8 27B
MedScribe (Vals AI)86.9%83.8%GPT-5.5
Vals Multimodal Index68.1%Not measuredOnly GPT-5.5 measured
IFBench75.9%Not measuredOnly GPT-5.5 measured
Long context reasoning (AA-LCR)84.3%82.0%GPT-5.5
DTBench96.0%88.0%GPT-5.5
LMCA54.3%41.4%GPT-5.5
Surface Evolver Bench88.1%45.0%GPT-5.5
Furniture assembly44.2%Not measuredOnly GPT-5.5 measured
ExploitBench47.4%Not measuredOnly GPT-5.5 measured
CL-bench Life22.2%Not measuredOnly GPT-5.5 measured
LMArena Text14711440GPT-5.5
LMArena Text: instruction following14781436GPT-5.5
LMArena Text: creative writing14551383GPT-5.5
LMArena Vision12971269GPT-5.5
LMArena Text: English14701448GPT-5.5
LMArena Text: non-English14671428GPT-5.5
LMArena Text: Chinese15341501GPT-5.5
LMArena Text: German14801443GPT-5.5
LMArena Text: Japanese14991387GPT-5.5
LMArena Text: Korean14601384GPT-5.5
LMArena Text: French14851465GPT-5.5
LMArena Text: Spanish14681449GPT-5.5
LMArena Text: Russian14731415GPT-5.5
LMArena Text: Polish14651439GPT-5.5
LMArena Text: multi-turn14761440GPT-5.5
LMArena Text: longer queries14841448GPT-5.5
LMArena Text: writing and language14661405GPT-5.5
LMArena Text: entertainment and media14491393GPT-5.5
LMArena Vision: English12921267GPT-5.5
LMArena Vision: Chinese13601326GPT-5.5
LMArena Vision: reading text13061281GPT-5.5
LMArena Vision: humor12931235GPT-5.5
LMArena Vision: recognizing things1231Not measuredOnly GPT-5.5 measured
LMArena Vision: creative writing13011237GPT-5.5
LMArena Document1486Not measuredOnly GPT-5.5 measured
Neither model has a result on the other 8 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-5.5 and Qwen3.8 27B
BenchmarkGPT-5.5Qwen3.8 27BLeader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

GPT-5.5$0.020Fits in one request
Qwen3.8 27B$0.0007Fits in one request

Qwen3.8 27B costs less

Repository review

50K input + 3K output tokens

GPT-5.5$0.340Fits in one request
Qwen3.8 27B$0.0060Fits in one request

Qwen3.8 27B costs less

Agent loop

200K input + 10K output tokens

GPT-5.5$1.30Fits in one request
Qwen3.8 27B$0.021Fits in one request

Qwen3.8 27B costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of GPT-5.5 and Qwen3.8 27B
SpecificationOpenAI GPT-5.5Qwen3.8 27B
MakerOpenAIQwen
WeightsNo public weights listedOpen weights
ReleasedApril 24, 2026August 14, 2026
Context window1.05M1M
Input $/Mtok$5.00$0.04
Output $/Mtok$30.00$1.35
Coding index67.873.6
Independent measurements behind it22
Cost per solved task$1.65$1.26
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5.5 better than Qwen3.8 27B for coding?

Pick Qwen3.8 27B for coding: it scores 73.6 on the Vibe Coding Index against 67.8 for GPT-5.5. Pick GPT-5.5 when the context window is larger, 1.05M against 1M tokens, or when what matters is AA Coding Index or LiveCodeBench (Vals AI), where it leads.

Which is cheaper, GPT-5.5 or Qwen3.8 27B?

GPT-5.5 costs $5.00 per million input tokens and $30.00 per million output tokens; Qwen3.8 27B costs $0.04 and $1.35. Per solved SWE-bench Verified task, GPT-5.5 costs $1.65 and Qwen3.8 27B $1.26.

Which has the bigger context window, GPT-5.5 or Qwen3.8 27B?

GPT-5.5, with 1.05M tokens against 1M.

How many benchmarks is the GPT-5.5 vs Qwen3.8 27B comparison based on?

82 benchmarks have a result for both models as of October 21, 2026. GPT-5.5's coding index rests on 2 independent benchmarks and Qwen3.8 27B's on 2. Where only one model has a result, the page says so instead of naming a winner.

Compare next

Embed this comparison on your site

Free under CC BY 4.0, while the source link stays visible. Preview the frame · Terms and other embeds

<iframe src="https://llmbenchmark.ai/embed/compare/gpt-5-5-vs-qwen3-8-27b" title="GPT-5.5 vs Qwen3.8 27B, LLMBenchmark.ai" width="100%" height="360" style="border:0;max-width:100%" loading="lazy"></iframe>
<p style="font:13px/1.4 sans-serif;margin:6px 0 0">Source: <a href="https://llmbenchmark.ai/compare/gpt-5-5-vs-qwen3-8-27b">LLMBenchmark.ai, GPT-5.5 vs Qwen3.8 27B</a></p>