Enter opens the highlighted result. Esc closes.

GLM 5.2 vs Qwen3 235B A22B Thinking 2507

GLM 5.2 leads Qwen3 235B A22B Thinking 2507 on the Artificial Analysis Intelligence Index, 33.7 to 12.7, and Qwen3 235B A22B Thinking 2507 costs 1.3x less per output token ($3.50 vs $4.40 per million). Updated October 22, 2026; 44 benchmarks have a result for both.

Replace GLM 5.2 Replace Qwen3 235B A22B Thinking 2507
GLM 5.2Z.ai
Coding
61.1
Overall
62.5
Output $/M
$4.40
42
Qwen3 235B A22B Thinking 2507Qwen
Coding
No data
Overall
52.9
Output $/M
$3.50
2
vs boards won of 44 shared

GLM 5.2 or Qwen3 235B A22B Thinking 2507: which to pick

Only GLM 5.2 has a coding score, 61.1 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding.

Pick Qwen3 235B A22B Thinking 2507 for output tokens 1.3x cheaper ($3.50 vs $4.40 per million) and work measured by OTIS Mock AIME or SimpleQA Verified, where it leads.

Model AGLM 5.2Z.ai61.1/1004 independent benchmarks$0.86 per solved taskModel BQwen3 235B A22B Thinking 2507QwenNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
44
GLM 5.2 only
54
Qwen3 235B A22B Thinking 2507 only
2
Like-for-like categories
8 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
GLM 5.2 61.1 from 4 benchmarks, Qwen3 235B A22B Thinking 2507 not benchmarked.
Widest category
Agentic work, 44.5 points apart on average over 2 shared benchmarks.
Widest single gap
Vending-Bench 2: GLM 5.2 $8,314 against Qwen3 235B A22B Thinking 2507 -$11, 52.0 points apart on a 0 to 100 scale.
Output price
GLM 5.2 $4.40 against Qwen3 235B A22B Thinking 2507 $3.50 per million output tokens, 1.3x apart.
Context window
GLM 5.2 takes 1.05M tokens, Qwen3 235B A22B Thinking 2507 128K.

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
CategoryGLM 5.2BarsQwen3 235B A22B Thinking 2507Ties
Coding40
Intelligence10
Agentic work20
Reasoning50
Math21
Knowledge41
Web design60
General180
All shared boards422
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
Coding benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
LMArena WebDev1604Not measuredOnly GLM 5.2 measured
AA Coding Index68.822.1GLM 5.2
LiveCodeBench (Vals AI)69.5%Not measuredOnly GLM 5.2 measured
SWE-bench Verified (Vals AI)82.8%Not measuredOnly GLM 5.2 measured
Vibe Code Bench (Vals AI)64.0%Not measuredOnly GLM 5.2 measured
Code migration (Vals AI)37.9%Not measuredOnly GLM 5.2 measured
ProgramBench (Vals AI)0.5%Not measuredOnly GLM 5.2 measured
Terminal-Bench 2.1 (Vals AI)67.8%Not measuredOnly GLM 5.2 measured
SkillsBench (Vals AI)45.1%Not measuredOnly GLM 5.2 measured
SciCode51.2%Not measuredOnly GLM 5.2 measured
Terminal-Bench Hard (Artificial Analysis run)50.8%Not measuredOnly GLM 5.2 measured
FrontierCode24.5%Not measuredOnly GLM 5.2 measured
WeirdML70.1%41.0%GLM 5.2
ALE-Bench1047.0Not measuredOnly GLM 5.2 measured
LMArena Text: coding14871424GLM 5.2
LMArena Text: software and IT14861431GLM 5.2
LMArena WebDev: React1617Not measuredOnly GLM 5.2 measured
LMArena WebDev: HTML1565Not measuredOnly GLM 5.2 measured
Neither model has a result on the other 21 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
Intelligence benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
AA Intelligence Index33.712.7GLM 5.2
Epoch Capabilities Index151.9Not measuredOnly GLM 5.2 measured
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
Agentic work benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
AA Agentic Index38.41.3GLM 5.2
Design Arena: agents building web apps1222Not measuredOnly GLM 5.2 measured
Design Arena: agents building full-stack apps1241Not measuredOnly GLM 5.2 measured
Design Arena: agents building mobile apps1159Not measuredOnly GLM 5.2 measured
Design Arena: agents building Android apps1178Not measuredOnly GLM 5.2 measured
Design Arena: agents building Godot games1142Not measuredOnly GLM 5.2 measured
Design Arena: agents building games1206Not measuredOnly GLM 5.2 measured
Design Arena: HTML slides1125Not measuredOnly GLM 5.2 measured
Design Arena: PowerPoint slides1170Not measuredOnly GLM 5.2 measured
SREBench (Vals AI)0.0%Not measuredOnly GLM 5.2 measured
Finance Agent (Vals AI)49.7%Not measuredOnly GLM 5.2 measured
Legal Research Bench (Vals AI)31.3%Not measuredOnly GLM 5.2 measured
Excel Modeling Benchmark (Vals AI)61.5%Not measuredOnly GLM 5.2 measured
τ²-Bench Telecom99.1%Not measuredOnly GLM 5.2 measured
APEX-Agents45.2%Not measuredOnly GLM 5.2 measured
PostTrainBench31.7%Not measuredOnly GLM 5.2 measured
Vending-Bench 2$8,314-$11GLM 5.2
Neither model has a result on the other 24 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
Reasoning benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
GPQA Diamond85.6%Not measuredOnly GLM 5.2 measured
GPQA Diamond (Artificial Analysis run)89.5%Not measuredOnly GLM 5.2 measured
Humanity's Last Exam (Artificial Analysis run)41.1%Not measuredOnly GLM 5.2 measured
GPQA Diamond (Epoch AI run)91.9%80.1%GLM 5.2
ARC-AGI77.0%Not measuredOnly GLM 5.2 measured
ARC-AGI-222.8%Not measuredOnly GLM 5.2 measured
SimpleBench58.8%Not measuredOnly GLM 5.2 measured
Chess puzzles22.0%Not measuredOnly GLM 5.2 measured
Mystery game puzzles19.0%9.0%GLM 5.2
EBR-bench9.5%Not measuredOnly GLM 5.2 measured
LMArena Text: hard prompts14811416GLM 5.2
LMArena Text: expert14871463GLM 5.2
LMArena Text: hard prompts in English14811432GLM 5.2
Kagi LLM Benchmark62.6%Not measuredOnly GLM 5.2 measured
Neither model has a result on the other 8 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index21
Math benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
OTIS Mock AIME86.4%86.7%Qwen3 235B A22B Thinking 2507
FrontierMath, tiers 1 to 359.2%Not measuredOnly GLM 5.2 measured
FrontierMath, tier 429.3%Not measuredOnly GLM 5.2 measured
ProofBench35.0%Not measuredOnly GLM 5.2 measured
LMArena Text: math14831412GLM 5.2
LMArena Text: mathematical work14891435GLM 5.2
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index41
Knowledge benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
MMLU Pro86.7%Not measuredOnly GLM 5.2 measured
LegalBench (Vals AI)84.1%Not measuredOnly GLM 5.2 measured
CorpFin (Vals AI)66.1%Not measuredOnly GLM 5.2 measured
TaxEval (Vals AI)73.3%Not measuredOnly GLM 5.2 measured
MedCode (Vals AI)40.8%Not measuredOnly GLM 5.2 measured
SimpleQA Verified34.2%40.4%Qwen3 235B A22B Thinking 2507
LMArena Text: science14891433GLM 5.2
LMArena Text: medicine and healthcare14751456GLM 5.2
LMArena Text: legal and government14861415GLM 5.2
LMArena Text: business and finance14511413GLM 5.2
Neither model has a result on the other 8 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index60
Web design benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
Design Arena: websites12851052GLM 5.2
Design Arena: UI components1294942GLM 5.2
Design Arena: game dev1268956GLM 5.2
Design Arena: data viz1285950GLM 5.2
Design Arena: 3D1287997GLM 5.2
Design Arena: all code categories12871032GLM 5.2
Design Arena: SVG1213Not measuredOnly GLM 5.2 measured
Design Arena: ASCII art1191Not measuredOnly GLM 5.2 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index180
General benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
MedScribe (Vals AI)83.5%Not measuredOnly GLM 5.2 measured
IFBench73.3%Not measuredOnly GLM 5.2 measured
Long context reasoning (AA-LCR)78.3%Not measuredOnly GLM 5.2 measured
DTBench93.6%80.3%GLM 5.2
LMCA45.8%29.3%GLM 5.2
Fiction.LiveBenchNot measured75.0%Only Qwen3 235B A22B Thinking 2507 measured
Creative writing (Lech Mazur)Not measured82.4%Only Qwen3 235B A22B Thinking 2507 measured
Surface Evolver Bench55.6%Not measuredOnly GLM 5.2 measured
LMArena Text14701415GLM 5.2
LMArena Text: instruction following14661386GLM 5.2
LMArena Text: creative writing14631384GLM 5.2
LMArena Text: English14751429GLM 5.2
LMArena Text: non-English14591397GLM 5.2
LMArena Text: Chinese15221481GLM 5.2
LMArena Text: German14701408GLM 5.2
LMArena Text: Japanese14501386GLM 5.2
LMArena Text: Korean14401375GLM 5.2
LMArena Text: French1479Not measuredOnly GLM 5.2 measured
LMArena Text: Spanish14781389GLM 5.2
LMArena Text: Russian14681401GLM 5.2
LMArena Text: Polish14561390GLM 5.2
LMArena Text: multi-turn14691406GLM 5.2
LMArena Text: longer queries14801400GLM 5.2
LMArena Text: writing and language14631388GLM 5.2
LMArena Text: entertainment and media14501370GLM 5.2
Neither model has a result on the other 21 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GLM 5.2 and Qwen3 235B A22B Thinking 2507
BenchmarkGLM 5.2Qwen3 235B A22B Thinking 2507Leader
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

GLM 5.2$0.0036Fits in one request
Qwen3 235B A22B Thinking 2507$0.0022Fits in one request

Qwen3 235B A22B Thinking 2507 costs less

Repository review

50K input + 3K output tokens

GLM 5.2$0.083Fits in one request
Qwen3 235B A22B Thinking 2507$0.033Fits in one request

Qwen3 235B A22B Thinking 2507 costs less

Agent loop

200K input + 10K output tokens

GLM 5.2$0.324Fits in one request
Qwen3 235B A22B Thinking 2507$0.125Exceeds the context window

Qwen3 235B A22B Thinking 2507 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 GLM 5.2 and Qwen3 235B A22B Thinking 2507
SpecificationZ.ai GLM 5.2Qwen3 235B A22B Thinking 2507
MakerZ.aiQwen
WeightsOpen weightsOpen weights
ReleasedJune 16, 2026July 25, 2025
Context window1.05M128K
Input $/Mtok$1.40$0.45
Output $/Mtok$4.40$3.50
Coding index61.1Not benchmarked
Independent measurements behind it40
Cost per solved task$0.86Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GLM 5.2 better than Qwen3 235B A22B Thinking 2507 for coding?

Only GLM 5.2 has a coding score, 61.1 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding. Pick Qwen3 235B A22B Thinking 2507 for output tokens 1.3x cheaper ($3.50 vs $4.40 per million) and work measured by OTIS Mock AIME or SimpleQA Verified, where it leads.

Which is cheaper, GLM 5.2 or Qwen3 235B A22B Thinking 2507?

GLM 5.2 costs $1.40 per million input tokens and $4.40 per million output tokens; Qwen3 235B A22B Thinking 2507 costs $0.45 and $3.50. Cost per solved task is known for GLM 5.2 only.

Which has the bigger context window, GLM 5.2 or Qwen3 235B A22B Thinking 2507?

GLM 5.2, with 1.05M tokens against 128K.

How many benchmarks is the GLM 5.2 vs Qwen3 235B A22B Thinking 2507 comparison based on?

44 benchmarks have a result for both models as of October 22, 2026. GLM 5.2's coding index rests on 4 independent benchmarks and Qwen3 235B A22B Thinking 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next