Enter opens the highlighted result. Esc closes.

GPT-5.2-Codex vs Qwen2.5 72B Instruct

Updated October 21, 2026. 6 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace GPT-5.2-Codex Replace Qwen2.5 72B Instruct
GPT-5.2-CodexOpenAI
Coding
45.4
Overall
28.5
Output $/M
$14.00
6
Qwen2.5 72B InstructQwen
Coding
No data
Overall
7.7
Output $/M
$0.40
0
vs boards won of 6 shared

Decision reading

Only GPT-5.2-Codex has a coding score, 45.4 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding.

Qwen2.5 72B Instruct has the edge where output tokens cost 35.0x less.

Model AGPT-5.2-CodexOpenAI45.4/1003 independent benchmarks$0.43 per solved taskModel BQwen2.5 72B InstructQwenNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
6
GPT-5.2-Codex only
16
Qwen2.5 72B Instruct only
47
Like-for-like categories
5 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-5.2-Codex 45.4 from 3 benchmarks, Qwen2.5 72B Instruct not benchmarked.
Widest category
Agentic work, 57.6 points apart on average over 1 shared benchmark.
Widest single gap
τ²-Bench Telecom: GPT-5.2-Codex 92.1% against Qwen2.5 72B Instruct 34.5%, 57.6 points apart on a 0 to 100 scale.
Output price
GPT-5.2-Codex $14.00 against Qwen2.5 72B Instruct $0.40 per million output tokens, 35.0x apart.
Context window
GPT-5.2-Codex takes 400K 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-5.2-CodexBarsQwen2.5 72B InstructTies
Coding10
Intelligence10
Agentic work10
Reasoning20
General10
All shared boards60
CodingLike-for-likeSix of these boards feed the Vibe Coding Index10
Coding benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
SWE-bench Verified72.8%Not measuredOnly GPT-5.2-Codex measured
SWE-bench Multilingual66.3%Not measuredOnly GPT-5.2-Codex measured
LMArena WebDev1339Not measuredOnly GPT-5.2-Codex measured
LiveCodeBench (Vals AI)88.0%Not measuredOnly GPT-5.2-Codex measured
SWE-bench Verified (Vals AI)72.4%Not measuredOnly GPT-5.2-Codex measured
Vibe Code Bench (Vals AI)37.9%Not measuredOnly GPT-5.2-Codex measured
LiveCodeBench (Artificial Analysis run)Not measured27.6%Only Qwen2.5 72B Instruct measured
Terminal-Bench Hard (Artificial Analysis run)37.1%4.5%GPT-5.2-Codex
Terminal-Bench66.5%Not measuredOnly GPT-5.2-Codex measured
WeirdMLNot measured16.0%Only Qwen2.5 72B Instruct measured
ALE-Bench1299.9Not measuredOnly GPT-5.2-Codex measured
LMArena Text: codingNot measured1292Only Qwen2.5 72B Instruct measured
LMArena Text: software and ITNot measured1292Only Qwen2.5 72B Instruct measured
LMArena WebDev: React1329Not measuredOnly GPT-5.2-Codex measured
LMArena WebDev: HTML1348Not measuredOnly GPT-5.2-Codex measured
Neither model has a result on the other 23 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
Intelligence benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
AA Intelligence Index28.57.7GPT-5.2-Codex
Epoch Capabilities IndexNot measured129.0Only Qwen2.5 72B Instruct measured
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
Agentic work benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
Design Arena: agents building web apps1051Not measuredOnly GPT-5.2-Codex measured
Design Arena: agents building full-stack apps993Not measuredOnly GPT-5.2-Codex measured
Design Arena: agents building mobile apps1089Not measuredOnly GPT-5.2-Codex measured
Design Arena: agents building Android apps1176Not measuredOnly GPT-5.2-Codex measured
Design Arena: agents building Godot games1142Not measuredOnly GPT-5.2-Codex measured
τ²-Bench Telecom92.1%34.5%GPT-5.2-Codex
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 30 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index20
Reasoning benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
GPQA Diamond (Artificial Analysis run)89.9%49.1%GPT-5.2-Codex
Humanity's Last Exam (Artificial Analysis run)35.7%3.6%GPT-5.2-Codex
GPQA Diamond (Epoch AI run)Not measured49.1%Only Qwen2.5 72B Instruct measured
BIG-Bench HardNot measured79.8%Only Qwen2.5 72B Instruct measured
LMArena Text: hard promptsNot measured1271Only Qwen2.5 72B Instruct measured
LMArena Text: expertNot measured1245Only Qwen2.5 72B Instruct measured
LMArena Text: hard prompts in EnglishNot measured1279Only Qwen2.5 72B Instruct measured
Neither model has a result on the other 14 boards here
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
Math benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 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
LMArena Text: mathNot measured1283Only Qwen2.5 72B Instruct measured
LMArena Text: mathematical workNot measured1290Only Qwen2.5 72B Instruct measured
Neither model has a result on the other 7 boards here
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
Knowledge benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
MMLU Pro (Artificial Analysis run)Not measured72.0%Only Qwen2.5 72B Instruct measured
MMLUNot measured85.3%Only Qwen2.5 72B Instruct measured
AI2 Reasoning ChallengeNot measured94.5%Only Qwen2.5 72B Instruct measured
LMArena Text: scienceNot measured1275Only Qwen2.5 72B Instruct measured
LMArena Text: medicine and healthcareNot measured1258Only Qwen2.5 72B Instruct measured
LMArena Text: legal and governmentNot measured1288Only Qwen2.5 72B Instruct measured
LMArena Text: business and financeNot measured1253Only Qwen2.5 72B Instruct measured
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-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index10
General benchmarks for GPT-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 72B InstructLeader
IFBench77.6%36.9%GPT-5.2-Codex
Long context reasoning (AA-LCR)82.3%Not measuredOnly GPT-5.2-Codex measured
DTBenchNot measured62.9%Only Qwen2.5 72B Instruct measured
LMCANot measured13.4%Only Qwen2.5 72B Instruct measured
HellaSwagNot measured84.8%Only Qwen2.5 72B Instruct measured
WinoGrandeNot measured82.3%Only Qwen2.5 72B Instruct measured
PIQANot measured82.6%Only Qwen2.5 72B Instruct measured
LMArena TextNot measured1269Only Qwen2.5 72B Instruct measured
LMArena Text: instruction followingNot measured1254Only Qwen2.5 72B Instruct measured
LMArena Text: creative writingNot measured1221Only Qwen2.5 72B Instruct measured
LMArena Text: EnglishNot measured1283Only Qwen2.5 72B Instruct measured
LMArena Text: non-EnglishNot measured1252Only Qwen2.5 72B Instruct measured
LMArena Text: ChineseNot measured1273Only Qwen2.5 72B Instruct measured
LMArena Text: GermanNot measured1234Only Qwen2.5 72B Instruct measured
LMArena Text: JapaneseNot measured1180Only Qwen2.5 72B Instruct measured
LMArena Text: KoreanNot measured1188Only Qwen2.5 72B Instruct measured
LMArena Text: FrenchNot measured1280Only Qwen2.5 72B Instruct measured
LMArena Text: SpanishNot measured1255Only Qwen2.5 72B Instruct measured
LMArena Text: RussianNot measured1264Only Qwen2.5 72B Instruct measured
LMArena Text: multi-turnNot measured1272Only Qwen2.5 72B Instruct measured
LMArena Text: longer queriesNot measured1281Only Qwen2.5 72B Instruct measured
LMArena Text: writing and languageNot measured1247Only Qwen2.5 72B Instruct measured
LMArena Text: entertainment and mediaNot measured1211Only Qwen2.5 72B Instruct measured
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-5.2-Codex and Qwen2.5 72B Instruct
BenchmarkGPT-5.2-CodexQwen2.5 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-5.2-Codex$0.0088Fits 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-5.2-Codex$0.130Fits in one request
Qwen2.5 72B Instruct$0.019Exceeds the context window

Qwen2.5 72B Instruct costs less

Agent loop

200K input + 10K output tokens

GPT-5.2-Codex$0.490Fits in one request
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-5.2-Codex and Qwen2.5 72B Instruct
SpecificationOpenAI GPT-5.2-CodexQwen2.5 72B Instruct
MakerOpenAIQwen
WeightsNo public weights listedOpen weights
ReleasedJanuary 14, 2026September 19, 2024
Context window400K33K
Input $/Mtok$1.75$0.36
Output $/Mtok$14.00$0.40
Coding index45.4Not benchmarked
Independent measurements behind it30
Cost per solved task$0.43Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5.2-Codex better than Qwen2.5 72B Instruct for coding?

Only GPT-5.2-Codex has a coding score, 45.4 on the Vibe Coding Index; Qwen2.5 72B Instruct has none, so the two cannot be ranked on coding. Qwen2.5 72B Instruct has the edge where output tokens cost 35.0x less.

Which is cheaper, GPT-5.2-Codex or Qwen2.5 72B Instruct?

GPT-5.2-Codex costs $1.75 per million input tokens and $14.00 per million output tokens; Qwen2.5 72B Instruct costs $0.36 and $0.40. Cost per solved task is known for GPT-5.2-Codex only.

Which has the bigger context window, GPT-5.2-Codex or Qwen2.5 72B Instruct?

GPT-5.2-Codex, with 400K tokens against 33K.

How many benchmarks is the GPT-5.2-Codex vs Qwen2.5 72B Instruct comparison based on?

6 benchmarks have a result for both models as of October 21, 2026. GPT-5.2-Codex's coding index rests on 3 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