DeepSeek V3.1 vs GLM 4.5 Air
DeepSeek V3.1 leads GLM 4.5 Air on the Artificial Analysis Intelligence Index, 13.7 to 11.1, and GLM 4.5 Air costs 1.4x less per output token ($1.10 vs $1.50 per million). Updated October 22, 2026; 9 benchmarks have a result for both.
DeepSeek V3.1 or GLM 4.5 Air: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick DeepSeek V3.1 for a larger context window (164K against 131K tokens) and work measured by Design Arena: websites or Design Arena: UI components, where it leads.
- Shared results
- 9
- DeepSeek V3.1 only
- 2
- GLM 4.5 Air only
- 39
- Like-for-like categories
- 3 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
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
DeepSeek V3.1
DeepSeek V3.1 leads, 13.7 against 11.1.
Same board, both measured
Web design
Design Arena: websites
GLM 4.5 Air
GLM 4.5 Air leads, 1146 against 1122.
Same board, both measured
Long documents
Listed context window
DeepSeek V3.1
DeepSeek V3.1 takes 164K tokens against 131K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GLM 4.5 Air
DeepSeek V3.1 $0.0013, GLM 4.5 Air $0.0008 per turn.
List prices
At a glance
- Coding index
- DeepSeek V3.1 not benchmarked, GLM 4.5 Air not benchmarked.
- Widest category
- Reasoning, 10.2 points apart on average over 1 shared benchmark.
- Widest single gap
- Design Arena: SVG: DeepSeek V3.1 967 against GLM 4.5 Air 1071, 13.0 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3.1 $1.50 against GLM 4.5 Air $1.10 per million output tokens, 1.4x apart.
- Context window
- DeepSeek V3.1 takes 164K tokens, GLM 4.5 Air 131K.
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 V3.1 | Bars | GLM 4.5 Air | Ties |
|---|---|---|---|---|
| Intelligence | 1 | 0 | ||
| Reasoning | 1 | 0 | ||
| Web design | 0 | 7 | ||
| All shared boards | 2 | 7 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index5 boards, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| LiveCodeBench (Artificial Analysis run) | Not measured | 68.4% | Only GLM 4.5 Air measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 20.5% | Only GLM 4.5 Air measured |
| GSO-Bench | Not measured | 2.9% | Only GLM 4.5 Air measured |
| LMArena Text: coding | Not measured | 1397 | Only GLM 4.5 Air measured |
| LMArena Text: software and IT | Not measured | 1401 | Only GLM 4.5 Air measured |
| Neither model has a result on the other 34 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| AA Intelligence Index | 13.7 | 11.1 | DeepSeek V3.1 |
| Neither model has a result on the other 1 board here | |||
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index1 board, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| τ²-Bench Telecom | Not measured | 46.5% | Only GLM 4.5 Air measured |
| Neither model has a result on the other 40 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | Not measured | 73.3% | Only GLM 4.5 Air measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 7.0% | Only GLM 4.5 Air measured |
| LMArena Text: hard prompts | Not measured | 1379 | Only GLM 4.5 Air measured |
| LMArena Text: expert | Not measured | 1370 | Only GLM 4.5 Air measured |
| LMArena Text: hard prompts in English | Not measured | 1388 | Only GLM 4.5 Air measured |
| Kagi LLM Benchmark | 53.2% | 43.0% | DeepSeek V3.1 |
| Neither model has a result on the other 16 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index4 boards, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 80.7 | Only GLM 4.5 Air measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 80.7% | Only GLM 4.5 Air measured |
| LMArena Text: math | Not measured | 1396 | Only GLM 4.5 Air measured |
| LMArena Text: mathematical work | Not measured | 1396 | Only GLM 4.5 Air measured |
| Neither model has a result on the other 9 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | Not measured | 81.5% | Only GLM 4.5 Air measured |
| LMArena Text: science | Not measured | 1407 | Only GLM 4.5 Air measured |
| LMArena Text: medicine and healthcare | Not measured | 1402 | Only GLM 4.5 Air measured |
| LMArena Text: legal and government | Not measured | 1382 | Only GLM 4.5 Air measured |
| LMArena Text: business and finance | Not measured | 1371 | Only GLM 4.5 Air measured |
| Neither model has a result on the other 13 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index07
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| Design Arena: websites | 1122 | 1146 | GLM 4.5 Air |
| Design Arena: UI components | 1085 | 1120 | GLM 4.5 Air |
| Design Arena: game dev | 1084 | 1094 | GLM 4.5 Air |
| Design Arena: data viz | 1096 | 1186 | GLM 4.5 Air |
| Design Arena: 3D | 1077 | 1122 | GLM 4.5 Air |
| Design Arena: all code categories | 1112 | 1137 | GLM 4.5 Air |
| Design Arena: SVG | 967 | 1071 | GLM 4.5 Air |
| Neither model has a result on the other 2 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index21 boards, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | Leader |
|---|---|---|---|
| IFBench | Not measured | 37.6% | Only GLM 4.5 Air measured |
| Long context reasoning (AA-LCR) | Not measured | 46.7% | Only GLM 4.5 Air measured |
| DTBench | 82.7% | Not measured | Only DeepSeek V3.1 measured |
| LMCA | 24.3% | Not measured | Only DeepSeek V3.1 measured |
| LMArena Text | Not measured | 1384 | Only GLM 4.5 Air measured |
| LMArena Text: instruction following | Not measured | 1354 | Only GLM 4.5 Air measured |
| LMArena Text: creative writing | Not measured | 1343 | Only GLM 4.5 Air measured |
| LMArena Text: English | Not measured | 1397 | Only GLM 4.5 Air measured |
| LMArena Text: non-English | Not measured | 1366 | Only GLM 4.5 Air measured |
| LMArena Text: Chinese | Not measured | 1426 | Only GLM 4.5 Air measured |
| LMArena Text: German | Not measured | 1377 | Only GLM 4.5 Air measured |
| LMArena Text: Japanese | Not measured | 1348 | Only GLM 4.5 Air measured |
| LMArena Text: Korean | Not measured | 1308 | Only GLM 4.5 Air measured |
| LMArena Text: French | Not measured | 1400 | Only GLM 4.5 Air measured |
| LMArena Text: Spanish | Not measured | 1385 | Only GLM 4.5 Air measured |
| LMArena Text: Russian | Not measured | 1373 | Only GLM 4.5 Air measured |
| LMArena Text: Polish | Not measured | 1368 | Only GLM 4.5 Air measured |
| LMArena Text: multi-turn | Not measured | 1371 | Only GLM 4.5 Air measured |
| LMArena Text: longer queries | Not measured | 1366 | Only GLM 4.5 Air measured |
| LMArena Text: writing and language | Not measured | 1354 | Only GLM 4.5 Air measured |
| LMArena Text: entertainment and media | Not measured | 1339 | Only GLM 4.5 Air measured |
| Neither model has a result on the other 25 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | DeepSeek V3.1 | GLM 4.5 Air | 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
GLM 4.5 Air costs less
Repository review
50K input + 3K output tokens
GLM 4.5 Air costs less
Agent loop
200K input + 10K output tokens
GLM 4.5 Air 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 V3.1 | Z.ai GLM 4.5 Air |
|---|---|---|
| Maker | DeepSeek | Z.ai |
| Weights | Open weights | Open weights |
| Released | August 21, 2025 | July 25, 2025 |
| Context window | 164K | 131K |
| Input $/Mtok | $0.55 | $0.20 |
| Output $/Mtok | $1.50 | $1.10 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is DeepSeek V3.1 better than GLM 4.5 Air for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick DeepSeek V3.1 for a larger context window (164K against 131K tokens) and work measured by Design Arena: websites or Design Arena: UI components, where it leads.
Which is cheaper, DeepSeek V3.1 or GLM 4.5 Air?
DeepSeek V3.1 costs $0.55 per million input tokens and $1.50 per million output tokens; GLM 4.5 Air costs $0.20 and $1.10. Cost per solved task is known for neither of them.
Which has the bigger context window, DeepSeek V3.1 or GLM 4.5 Air?
DeepSeek V3.1, with 164K tokens against 131K.
How many benchmarks is the DeepSeek V3.1 vs GLM 4.5 Air comparison based on?
9 benchmarks have a result for both models as of October 22, 2026. DeepSeek V3.1's coding index rests on 0 independent benchmarks and GLM 4.5 Air's on 0. Where only one model has a result, the page says so instead of naming a winner.