LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

GPT-4 vs Qwen3.5-27B

Updated October 21, 2026. 33 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace GPT-4 Replace Qwen3.5-27B
GPT-4OpenAI
Coding
22.4
Overall
6.7
Output $/M
$60.00
0
Qwen3.5-27BQwen
Coding
22.6
Overall
22.9
Output $/M
$2.16
33
vs boards won of 33 shared

Decision reading

GPT-4 and Qwen3.5-27B are level for coding, 22.4 against 22.6 on the Vibe Coding Index, so price and context decide.

GPT-4 is neither cheaper nor larger in context, and leads no benchmark both models were measured on. Pick Qwen3.5-27B when output tokens cost 27.8x less, or when the context window is larger, 262K against 8K tokens.

Model AGPT-4OpenAI22.4/1001 independent benchmarkNo cost per solved taskModel BQwen3.5-27BQwen22.6/1001 independent benchmarkNo cost per solved task
Shared results
33
GPT-4 only
12
Qwen3.5-27B only
20
Like-for-like categories
6 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-4 22.4 from 1 benchmark, Qwen3.5-27B 22.6 from 1 benchmark.
Widest category
Reasoning, 42.4 points apart on average over 4 shared benchmarks.
Widest single gap
GPQA Diamond (Artificial Analysis run): GPT-4 34.9% against Qwen3.5-27B 85.8%, 50.9 points apart on a 0 to 100 scale.
Output price
GPT-4 $60.00 against Qwen3.5-27B $2.16 per million output tokens, 27.8x apart.
Context window
GPT-4 takes 8K tokens, Qwen3.5-27B 262K.

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-4BarsQwen3.5-27BTies
Coding02
Intelligence01
Reasoning04
Math02
Knowledge04
General020
All shared boards033
CodingLike-for-likeFour of these boards feed the Vibe Coding Index02
Coding benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
SWE-bench Verified22.4%Not measuredOnly GPT-4 measured
SWE-bench Lite39.7%Not measuredOnly GPT-4 measured
SWE-bench (full test set)12.5%Not measuredOnly GPT-4 measured
LMArena WebDevNot measured1358Only Qwen3.5-27B measured
AA Coding Index13.1Not measuredOnly GPT-4 measured
Terminal-Bench Hard (Artificial Analysis run)Not measured32.6%Only Qwen3.5-27B measured
WeirdMLNot measured39.5%Only Qwen3.5-27B measured
ALE-BenchNot measured349.4Only Qwen3.5-27B measured
LMArena Text: coding12091427Qwen3.5-27B
LMArena Text: software and IT11951431Qwen3.5-27B
LMArena WebDev: ReactNot measured1346Only Qwen3.5-27B measured
LMArena WebDev: HTMLNot measured1386Only Qwen3.5-27B measured
Neither model has a result on the other 26 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
Intelligence benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
AA Intelligence Index6.722.9Qwen3.5-27B
Epoch Capabilities Index125.7Not measuredOnly GPT-4 measured
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
Agentic work benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
τ²-Bench TelecomNot measured93.9%Only Qwen3.5-27B measured
Vending-Bench 2Not measured$202Only Qwen3.5-27B measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index04
Reasoning benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
GPQA Diamond (Artificial Analysis run)34.9%85.8%Qwen3.5-27B
Humanity's Last Exam (Artificial Analysis run)Not measured23.9%Only Qwen3.5-27B measured
BIG-Bench Hard75.1%Not measuredOnly GPT-4 measured
LMArena Text: hard prompts12001415Qwen3.5-27B
LMArena Text: expert11491427Qwen3.5-27B
LMArena Text: hard prompts in English12091429Qwen3.5-27B
LMArena Vision: diagramsNot measured1249Only Qwen3.5-27B measured
LMArena Vision: homeworkNot measured1256Only Qwen3.5-27B measured
Neither model has a result on the other 13 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index02
Math benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
GSM8K92.0%Not measuredOnly GPT-4 measured
LMArena Text: math12301429Qwen3.5-27B
LMArena Text: mathematical work12281429Qwen3.5-27B
Neither model has a result on the other 10 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index04
Knowledge benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
MMLU Pro (Artificial Analysis run)56.2%Not measuredOnly GPT-4 measured
MMLU86.4%Not measuredOnly GPT-4 measured
AI2 Reasoning Challenge96.3%Not measuredOnly GPT-4 measured
LMArena Text: science11771437Qwen3.5-27B
LMArena Text: medicine and healthcare11451425Qwen3.5-27B
LMArena Text: legal and government11911405Qwen3.5-27B
LMArena Text: business and finance11451399Qwen3.5-27B
Neither model has a result on the other 10 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index020
General benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
IFBench33.2%75.6%Qwen3.5-27B
Long context reasoning (AA-LCR)Not measured77.7%Only Qwen3.5-27B measured
DTBench62.7%82.4%Qwen3.5-27B
LMCA17.1%34.0%Qwen3.5-27B
HellaSwag95.3%Not measuredOnly GPT-4 measured
WinoGrande87.5%Not measuredOnly GPT-4 measured
LMArena Text12061409Qwen3.5-27B
LMArena Text: instruction following12011393Qwen3.5-27B
LMArena Text: creative writing11911362Qwen3.5-27B
LMArena VisionNot measured1241Only Qwen3.5-27B measured
LMArena Text: English12201425Qwen3.5-27B
LMArena Text: non-English11941390Qwen3.5-27B
LMArena Text: Chinese11841478Qwen3.5-27B
LMArena Text: German11981392Qwen3.5-27B
LMArena Text: Japanese11371346Qwen3.5-27B
LMArena Text: Korean10871359Qwen3.5-27B
LMArena Text: French12201410Qwen3.5-27B
LMArena Text: Spanish11971407Qwen3.5-27B
LMArena Text: Russian11951390Qwen3.5-27B
LMArena Text: Polish12651375Qwen3.5-27B
LMArena Text: multi-turn12061410Qwen3.5-27B
LMArena Text: longer queries11901413Qwen3.5-27B
LMArena Text: writing and language12191380Qwen3.5-27B
LMArena Text: entertainment and media11841363Qwen3.5-27B
LMArena Vision: EnglishNot measured1246Only Qwen3.5-27B measured
LMArena Vision: ChineseNot measured1301Only Qwen3.5-27B measured
LMArena Vision: reading textNot measured1246Only Qwen3.5-27B measured
LMArena Vision: humorNot measured1227Only Qwen3.5-27B measured
LMArena Vision: captioningNot measured1170Only Qwen3.5-27B measured
LMArena Vision: recognizing thingsNot measured1210Only Qwen3.5-27B measured
LMArena Vision: creative writingNot measured1224Only Qwen3.5-27B measured
Neither model has a result on the other 15 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-4 and Qwen3.5-27B
BenchmarkGPT-4Qwen3.5-27BLeader
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-4$0.060Fits in one request
Qwen3.5-27B$0.0014Fits in one request

Qwen3.5-27B costs less

Repository review

50K input + 3K output tokens

GPT-4$1.68Exceeds the context window
Qwen3.5-27B$0.020Fits in one request

Qwen3.5-27B costs less

Agent loop

200K input + 10K output tokens

GPT-4$6.60Exceeds the context window
Qwen3.5-27B$0.076Fits in one request

Qwen3.5-27B 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-4 and Qwen3.5-27B
SpecificationOpenAI GPT-4Qwen3.5-27B
MakerOpenAIQwen
WeightsNo public weights listedOpen weights
ReleasedMay 28, 2023February 25, 2026
Context window8K262K
Input $/Mtok$30.00$0.27
Output $/Mtok$60.00$2.16
Coding index22.422.6
Independent measurements behind it11
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-4 better than Qwen3.5-27B for coding?

GPT-4 and Qwen3.5-27B are level for coding, 22.4 against 22.6 on the Vibe Coding Index, so price and context decide. GPT-4 is neither cheaper nor larger in context, and leads no benchmark both models were measured on. Pick Qwen3.5-27B when output tokens cost 27.8x less, or when the context window is larger, 262K against 8K tokens.

Which is cheaper, GPT-4 or Qwen3.5-27B?

GPT-4 costs $30.00 per million input tokens and $60.00 per million output tokens; Qwen3.5-27B costs $0.27 and $2.16. Cost per solved task is known for neither of them.

Which has the bigger context window, GPT-4 or Qwen3.5-27B?

Qwen3.5-27B, with 262K tokens against 8K.

How many benchmarks is the GPT-4 vs Qwen3.5-27B comparison based on?

33 benchmarks have a result for both models as of October 21, 2026. GPT-4's coding index rests on 1 independent benchmark and Qwen3.5-27B's on 1. Where only one model has a result, the page says so instead of naming a winner.

Compare next