DeepSeek V3.2 vs GLM 5
GLM 5 leads DeepSeek V3.2 on coding, 41.7 to 30.0 on the Vibe Coding Index, and DeepSeek V3.2 costs 3.6x less per output token ($0.88 vs $3.20 per million). Updated October 22, 2026; 65 benchmarks have a result for both.
DeepSeek V3.2 or GLM 5: which to pick
Pick GLM 5 for coding: it scores 41.7 on the Vibe Coding Index against 30.0 for DeepSeek V3.2.
Pick DeepSeek V3.2 for output tokens 3.6x cheaper ($0.88 vs $3.20 per million) and work measured by Epoch Capabilities Index or GPQA Diamond (Artificial Analysis run), where it leads.
- Shared results
- 65
- DeepSeek V3.2 only
- 10
- GLM 5 only
- 11
- 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.
Coding
Vibe Coding Index, six public coding benchmarks
GLM 5
GLM 5 scores 41.7 against 30.0.
3 benchmarks behind each score
Intelligence
AA Intelligence Index
GLM 5
GLM 5 leads, 27.9 against 21.5.
Same board, both measured
Reasoning
GPQA Diamond
GLM 5
GLM 5 leads, 83.3% against 80.3%.
Same board, both measured
Math
AIME
GLM 5
GLM 5 leads, 91.7% against 84.6%.
Same board, both measured
Knowledge
MMLU Pro
GLM 5
GLM 5 leads, 86.0% against 84.9%.
Same board, both measured
Web design
Design Arena: websites
GLM 5
GLM 5 leads, 1247 against 1175.
Same board, both measured
Long documents
Listed context window
GLM 5
GLM 5 takes 205K tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V3.2
DeepSeek V3.2 $0.0007, GLM 5 $0.0026 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
GLM 5
DeepSeek V3.2 $0.99, GLM 5 $0.63.
Published attempt cost and pass rate
At a glance
- Coding index
- DeepSeek V3.2 30.0 from 3 benchmarks, GLM 5 41.7 from 3 benchmarks.
- Widest category
- Agentic work, 14.4 points apart on average over 2 shared benchmarks.
- Widest single gap
- Vending-Bench 2: DeepSeek V3.2 $1,034 against GLM 5 $4,432, 21.2 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3.2 $0.88 against GLM 5 $3.20 per million output tokens, 3.6x apart.
- Context window
- DeepSeek V3.2 takes 164K tokens, GLM 5 205K.
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.
| Category | DeepSeek V3.2 | Bars | GLM 5 | Ties |
|---|---|---|---|---|
| Coding | 0 | 12 | ||
| Intelligence | 1 | 1 | ||
| Agentic work | 0 | 2 | ||
| Reasoning | 2 | 6 | ||
| Math | 0 | 3 | ||
| Knowledge | 1 | 9 | ||
| Web design | 0 | 8 | ||
| General | 0 | 20 | ||
| All shared boards | 4 | 61 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index012
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| SWE-bench Verified | 70.0% | 72.8% | GLM 5 |
| SWE-bench Multilingual | 59.0% | 69.7% | GLM 5 |
| LMArena WebDev | 1362 | 1435 | GLM 5 |
| AA Coding Index | 44.2 | Not measured | Only DeepSeek V3.2 measured |
| LiveCodeBench (Vals AI) | 80.7% | 81.9% | GLM 5 |
| SWE-bench Verified (Vals AI) | 67.6% | 71.4% | GLM 5 |
| Vibe Code Bench (Vals AI) | 5.1% | 23.4% | GLM 5 |
| LiveCodeBench (Artificial Analysis run) | 86.2% | Not measured | Only DeepSeek V3.2 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 35.6% | 43.2% | GLM 5 |
| Terminal-Bench | 39.6% | 52.4% | GLM 5 |
| WeirdML | Not measured | 48.2% | Only GLM 5 measured |
| ALE-Bench | Not measured | 765.6 | Only GLM 5 measured |
| LMArena Text: coding | 1453 | 1461 | GLM 5 |
| LMArena Text: software and IT | 1449 | 1460 | GLM 5 |
| LMArena WebDev: React | 1350 | 1424 | GLM 5 |
| LMArena WebDev: HTML | 1371 | 1450 | GLM 5 |
| Neither model has a result on the other 23 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index11
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| AA Intelligence Index | 21.5 | 27.9 | GLM 5 |
| Epoch Capabilities Index | 146.1 | 145.8 | DeepSeek V3.2 |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| Design Arena: agents building full-stack apps | Not measured | 1123 | Only GLM 5 measured |
| Design Arena: agents building mobile apps | Not measured | 1136 | Only GLM 5 measured |
| Design Arena: agents building Android apps | Not measured | 1147 | Only GLM 5 measured |
| Design Arena: agents building Godot games | Not measured | 1144 | Only GLM 5 measured |
| Design Arena: HTML slides | Not measured | 1098 | Only GLM 5 measured |
| τ²-Bench Telecom | 90.6% | 98.2% | GLM 5 |
| APEX-Agents | 21.3% | Not measured | Only DeepSeek V3.2 measured |
| Vending-Bench 2 | $1,034 | $4,432 | GLM 5 |
| Neither model has a result on the other 33 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index26
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| GPQA Diamond | 80.3% | 83.3% | GLM 5 |
| GPQA Diamond (Artificial Analysis run) | 84.0% | 82.0% | DeepSeek V3.2 |
| Humanity's Last Exam (Artificial Analysis run) | 24.6% | 29.3% | GLM 5 |
| GPQA Diamond (Epoch AI run) | Not measured | 87.8% | Only GLM 5 measured |
| ARC-AGI | 57.0% | 44.7% | DeepSeek V3.2 |
| ARC-AGI-2 | 4.0% | 4.9% | GLM 5 |
| SimpleBench | Not measured | 53.2% | Only GLM 5 measured |
| LMArena Text: hard prompts | 1434 | 1452 | GLM 5 |
| LMArena Text: expert | 1436 | 1454 | GLM 5 |
| LMArena Text: hard prompts in English | 1446 | 1460 | GLM 5 |
| Kagi LLM Benchmark | Not measured | 75.0% | Only GLM 5 measured |
| Neither model has a result on the other 11 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| AIME | 84.6% | 91.7% | GLM 5 |
| MGSM (Vals AI) | 86.0% | Not measured | Only DeepSeek V3.2 measured |
| AA Math Index | 92.0 | Not measured | Only DeepSeek V3.2 measured |
| AIME 2025 (Artificial Analysis run) | 92.0% | Not measured | Only DeepSeek V3.2 measured |
| OTIS Mock AIME | Not measured | 80.0% | Only GLM 5 measured |
| LMArena Text: math | 1435 | 1440 | GLM 5 |
| LMArena Text: mathematical work | 1440 | 1451 | GLM 5 |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index19
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| MMLU Pro | 84.9% | 86.0% | GLM 5 |
| LegalBench (Vals AI) | 76.1% | 84.1% | GLM 5 |
| CaseLaw (Vals AI) | 55.4% | 52.5% | DeepSeek V3.2 |
| CorpFin (Vals AI) | 51.0% | 62.9% | GLM 5 |
| TaxEval (Vals AI) | 68.2% | 70.0% | GLM 5 |
| MedQA (Vals AI) | 93.9% | 94.3% | GLM 5 |
| MMLU Pro (Artificial Analysis run) | 86.2% | Not measured | Only DeepSeek V3.2 measured |
| LMArena Text: science | 1446 | 1460 | GLM 5 |
| LMArena Text: medicine and healthcare | 1440 | 1455 | GLM 5 |
| LMArena Text: legal and government | 1425 | 1453 | GLM 5 |
| LMArena Text: business and finance | 1410 | 1430 | GLM 5 |
| Neither model has a result on the other 7 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index08
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| Design Arena: websites | 1175 | 1247 | GLM 5 |
| Design Arena: UI components | 1146 | 1227 | GLM 5 |
| Design Arena: game dev | 1131 | 1234 | GLM 5 |
| Design Arena: data viz | 1158 | 1228 | GLM 5 |
| Design Arena: 3D | 1132 | 1228 | GLM 5 |
| Design Arena: all code categories | 1163 | 1246 | GLM 5 |
| Design Arena: SVG | 1032 | 1168 | GLM 5 |
| Design Arena: ASCII art | 1083 | 1147 | GLM 5 |
| Neither model has a result on the other 1 board here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index020
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| IFBench | 60.7% | 72.3% | GLM 5 |
| Long context reasoning (AA-LCR) | 73.3% | 75.7% | GLM 5 |
| DTBench | 85.6% | Not measured | Only DeepSeek V3.2 measured |
| LMCA | 28.8% | Not measured | Only DeepSeek V3.2 measured |
| CL-bench | 12.4% | 18.7% | GLM 5 |
| CL-bench Life | 7.4% | Not measured | Only DeepSeek V3.2 measured |
| LMArena Text | 1425 | 1446 | GLM 5 |
| LMArena Text: instruction following | 1413 | 1428 | GLM 5 |
| LMArena Text: creative writing | 1400 | 1438 | GLM 5 |
| LMArena Text: English | 1439 | 1458 | GLM 5 |
| LMArena Text: non-English | 1409 | 1430 | GLM 5 |
| LMArena Text: Chinese | 1455 | 1511 | GLM 5 |
| LMArena Text: German | 1413 | 1447 | GLM 5 |
| LMArena Text: Japanese | 1374 | 1416 | GLM 5 |
| LMArena Text: Korean | 1370 | 1423 | GLM 5 |
| LMArena Text: French | 1429 | 1456 | GLM 5 |
| LMArena Text: Spanish | 1424 | 1454 | GLM 5 |
| LMArena Text: Russian | 1424 | 1436 | GLM 5 |
| LMArena Text: Polish | 1419 | 1434 | GLM 5 |
| LMArena Text: multi-turn | 1427 | 1456 | GLM 5 |
| LMArena Text: longer queries | 1428 | 1446 | GLM 5 |
| LMArena Text: writing and language | 1407 | 1437 | GLM 5 |
| LMArena Text: entertainment and media | 1395 | 1426 | GLM 5 |
| 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
| Benchmark | DeepSeek V3.2 | GLM 5 | Leader |
|---|---|---|---|
| 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
DeepSeek V3.2 costs less
Repository review
50K input + 3K output tokens
DeepSeek V3.2 costs less
Agent loop
200K input + 10K output tokens
DeepSeek V3.2 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.
| Specification | DeepSeek V3.2 | Z.ai GLM 5 |
|---|---|---|
| Maker | DeepSeek | Z.ai |
| Weights | Open weights | Open weights |
| Released | December 1, 2025 | February 11, 2026 |
| Context window | 164K | 205K |
| Input $/Mtok | $0.29 | $1.00 |
| Output $/Mtok | $0.88 | $3.20 |
| Coding index | 30.0 | 41.7 |
| Independent measurements behind it | 3 | 3 |
| Cost per solved task | $0.99 | $0.63 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V3.2 better than GLM 5 for coding?
Pick GLM 5 for coding: it scores 41.7 on the Vibe Coding Index against 30.0 for DeepSeek V3.2. Pick DeepSeek V3.2 for output tokens 3.6x cheaper ($0.88 vs $3.20 per million) and work measured by Epoch Capabilities Index or GPQA Diamond (Artificial Analysis run), where it leads.
Which is cheaper, DeepSeek V3.2 or GLM 5?
DeepSeek V3.2 costs $0.29 per million input tokens and $0.88 per million output tokens; GLM 5 costs $1.00 and $3.20. Per solved SWE-bench Verified task, DeepSeek V3.2 costs $0.99 and GLM 5 $0.63.
Which has the bigger context window, DeepSeek V3.2 or GLM 5?
GLM 5, with 205K tokens against 164K.
How many benchmarks is the DeepSeek V3.2 vs GLM 5 comparison based on?
65 benchmarks have a result for both models as of October 22, 2026. DeepSeek V3.2's coding index rests on 3 independent benchmarks and GLM 5's on 3. Where only one model has a result, the page says so instead of naming a winner.