Enter opens the highlighted result. Esc closes.

GPT-4 vs Qwen2.5 72B Instruct

Qwen2.5 72B Instruct leads GPT-4 on the Artificial Analysis Intelligence Index, 7.7 to 6.7, and Qwen2.5 72B Instruct costs 150.0x less per output token ($0.40 vs $60.00 per million). Updated October 22, 2026; 39 benchmarks have a result for both.

Replace GPT-4 Replace Qwen2.5 72B Instruct
GPT-4OpenAI
Coding
No data
Overall
No data
Output $/M
$60.00
5
Qwen2.5 72B InstructQwen
Coding
No data
Overall
26.5
Output $/M
$0.40
34
vs boards won of 39 shared

GPT-4 or Qwen2.5 72B Instruct: which to pick

Neither model has a coding score yet, so the two cannot be ranked on coding.

Pick GPT-4 for work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Model AGPT-4OpenAINot benchmarkedNo coding benchmarkNo cost per solved taskModel BQwen2.5 72B InstructQwenNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
39
GPT-4 only
6
Qwen2.5 72B Instruct only
14
Like-for-like categories
6 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

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 not benchmarked, Qwen2.5 72B Instruct not benchmarked.
Widest category
Coding, 15.1 points apart on average over 2 shared benchmarks.
Widest single gap
LMArena Text: medicine and healthcare: GPT-4 1145 against Qwen2.5 72B Instruct 1258, 18.9 points apart on a 0 to 100 scale.
Output price
GPT-4 $60.00 against Qwen2.5 72B Instruct $0.40 per million output tokens, 150.0x apart.
Context window
GPT-4 takes 8K tokens, Qwen2.5 72B Instruct 33K.

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-4BarsQwen2.5 72B InstructTies
Coding02
Intelligence02
Reasoning05
Math02
Knowledge25
General318
All shared boards534
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
Coding benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
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
AA Coding Index13.1Not measuredOnly GPT-4 measured
LiveCodeBench (Artificial Analysis run)Not measured27.6%Only Qwen2.5 72B Instruct measured
Terminal-Bench Hard (Artificial Analysis run)Not measured4.5%Only Qwen2.5 72B Instruct measured
WeirdMLNot measured16.0%Only Qwen2.5 72B Instruct measured
LMArena Text: coding12091292Qwen2.5 72B Instruct
LMArena Text: software and IT11951292Qwen2.5 72B Instruct
Neither model has a result on the other 30 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index02
Intelligence benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
AA Intelligence Index6.77.7Qwen2.5 72B Instruct
Epoch Capabilities Index125.7129.0Qwen2.5 72B Instruct
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index4 boards, none shared
Agentic work benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
τ²-Bench TelecomNot measured34.5%Only Qwen2.5 72B Instruct measured
METR time horizonsNot measured35.8%Only Qwen2.5 72B Instruct measured
BALROGNot measured16.2%Only Qwen2.5 72B Instruct measured
TheAgentCompanyNot measured5.7%Only Qwen2.5 72B Instruct measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index05
Reasoning benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
GPQA Diamond (Artificial Analysis run)34.9%49.1%Qwen2.5 72B Instruct
Humanity's Last Exam (Artificial Analysis run)Not measured3.6%Only Qwen2.5 72B Instruct measured
GPQA Diamond (Epoch AI run)Not measured49.1%Only Qwen2.5 72B Instruct measured
BIG-Bench Hard75.1%79.8%Qwen2.5 72B Instruct
LMArena Text: hard prompts12001271Qwen2.5 72B Instruct
LMArena Text: expert11491245Qwen2.5 72B Instruct
LMArena Text: hard prompts in English12091279Qwen2.5 72B Instruct
Neither model has a result on the other 15 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index02
Math benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
AA Math IndexNot measured14.0Only Qwen2.5 72B Instruct measured
AIME 2025 (Artificial Analysis run)Not measured14.0%Only Qwen2.5 72B Instruct measured
OTIS Mock AIMENot measured8.1%Only Qwen2.5 72B Instruct measured
MATH level 5Not measured63.2%Only Qwen2.5 72B Instruct measured
GSM8K92.0%Not measuredOnly GPT-4 measured
LMArena Text: math12301283Qwen2.5 72B Instruct
LMArena Text: mathematical work12281290Qwen2.5 72B Instruct
Neither model has a result on the other 6 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index25
Knowledge benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
MMLU Pro (Artificial Analysis run)56.2%72.0%Qwen2.5 72B Instruct
MMLU86.4%85.3%GPT-4
AI2 Reasoning Challenge96.3%94.5%GPT-4
LMArena Text: science11771275Qwen2.5 72B Instruct
LMArena Text: medicine and healthcare11451258Qwen2.5 72B Instruct
LMArena Text: legal and government11911288Qwen2.5 72B Instruct
LMArena Text: business and finance11451253Qwen2.5 72B Instruct
Neither model has a result on the other 11 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index318
General benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
IFBench33.2%36.9%Qwen2.5 72B Instruct
DTBench62.7%62.9%Qwen2.5 72B Instruct
LMCA17.1%13.4%GPT-4
HellaSwag95.3%84.8%GPT-4
WinoGrande87.5%82.3%GPT-4
PIQANot measured82.6%Only Qwen2.5 72B Instruct measured
LMArena Text12061269Qwen2.5 72B Instruct
LMArena Text: instruction following12011254Qwen2.5 72B Instruct
LMArena Text: creative writing11911221Qwen2.5 72B Instruct
LMArena Text: English12201283Qwen2.5 72B Instruct
LMArena Text: non-English11941252Qwen2.5 72B Instruct
LMArena Text: Chinese11841273Qwen2.5 72B Instruct
LMArena Text: German11981234Qwen2.5 72B Instruct
LMArena Text: Japanese11371180Qwen2.5 72B Instruct
LMArena Text: Korean10871188Qwen2.5 72B Instruct
LMArena Text: French12201280Qwen2.5 72B Instruct
LMArena Text: Spanish11971255Qwen2.5 72B Instruct
LMArena Text: Russian11951264Qwen2.5 72B Instruct
LMArena Text: Polish1265Not measuredOnly GPT-4 measured
LMArena Text: multi-turn12061272Qwen2.5 72B Instruct
LMArena Text: longer queries11901281Qwen2.5 72B Instruct
LMArena Text: writing and language12191247Qwen2.5 72B Instruct
LMArena Text: entertainment and media11841211Qwen2.5 72B Instruct
Neither model has a result on the other 23 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-4 and Qwen2.5 72B Instruct
BenchmarkGPT-4Qwen2.5 72B InstructLeader
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
Qwen2.5 72B Instruct$0.0006Fits in one request

Qwen2.5 72B Instruct costs less

Repository review

50K input + 3K output tokens

GPT-4$1.68Exceeds the context window
Qwen2.5 72B Instruct$0.019Exceeds the context window

Qwen2.5 72B Instruct costs less

Agent loop

200K input + 10K output tokens

GPT-4$6.60Exceeds the context window
Qwen2.5 72B Instruct$0.076Exceeds the context window

Qwen2.5 72B Instruct 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 Qwen2.5 72B Instruct
SpecificationOpenAI GPT-4Qwen2.5 72B Instruct
MakerOpenAIQwen
WeightsNo public weights listedOpen weights
ReleasedMay 28, 2023September 19, 2024
Context window8K33K
Input $/Mtok$30.00$0.36
Output $/Mtok$60.00$0.40
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it00
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-4 better than Qwen2.5 72B Instruct for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick GPT-4 for work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Which is cheaper, GPT-4 or Qwen2.5 72B Instruct?

GPT-4 costs $30.00 per million input tokens and $60.00 per million output tokens; Qwen2.5 72B Instruct costs $0.36 and $0.40. Cost per solved task is known for neither of them.

Which has the bigger context window, GPT-4 or Qwen2.5 72B Instruct?

Qwen2.5 72B Instruct, with 33K tokens against 8K.

How many benchmarks is the GPT-4 vs Qwen2.5 72B Instruct comparison based on?

39 benchmarks have a result for both models as of October 22, 2026. GPT-4's coding index rests on 0 independent benchmarks and Qwen2.5 72B Instruct's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next