GLM 4.5 Air vs Mistral Medium 3
GLM 4.5 Air leads Mistral Medium 3 on the Artificial Analysis Intelligence Index, 11.1 to 9.0, and GLM 4.5 Air costs 1.8x less per output token ($1.10 vs $2.00 per million). Updated October 22, 2026; 45 benchmarks have a result for both.
GLM 4.5 Air or Mistral Medium 3: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick GLM 4.5 Air for output tokens 1.8x cheaper ($1.10 vs $2.00 per million) and work measured by LMArena Text: legal and government or IFBench, where it leads.
- Shared results
- 45
- GLM 4.5 Air only
- 3
- Mistral Medium 3 only
- 12
- 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
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
GLM 4.5 Air
GLM 4.5 Air leads, 11.1 against 9.0.
Same board, both measured
Web design
Design Arena: websites
GLM 4.5 Air
GLM 4.5 Air leads, 1146 against 1079.
Same board, both measured
Long documents
Listed context window
No call
Both take 131K tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
GLM 4.5 Air
GLM 4.5 Air $0.0008, Mistral Medium 3 $0.0014 per turn.
List prices
At a glance
- Coding index
- GLM 4.5 Air not benchmarked, Mistral Medium 3 not benchmarked.
- Widest category
- Math, 28.7 points apart on average over 4 shared benchmarks.
- Widest single gap
- AA Math Index: GLM 4.5 Air 80.7 against Mistral Medium 3 30.3, 50.4 points apart on a 0 to 100 scale.
- Output price
- GLM 4.5 Air $1.10 against Mistral Medium 3 $2.00 per million output tokens, 1.8x apart.
- Context window
- Both take 131K tokens.
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 | GLM 4.5 Air | Bars | Mistral Medium 3 | Ties |
|---|---|---|---|---|
| Coding | 4 | 0 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 5 | 0 | ||
| Math | 4 | 0 | ||
| Knowledge | 4 | 1 | ||
| Web design | 6 | 0 | ||
| General | 15 | 4 | ||
| All shared boards | 40 | 5 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index40
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| LiveCodeBench (Artificial Analysis run) | 68.4% | 40.0% | GLM 4.5 Air |
| Terminal-Bench Hard (Artificial Analysis run) | 20.5% | 3.8% | GLM 4.5 Air |
| GSO-Bench | 2.9% | Not measured | Only GLM 4.5 Air measured |
| LMArena Text: coding | 1397 | 1385 | GLM 4.5 Air |
| LMArena Text: software and IT | 1401 | 1385 | GLM 4.5 Air |
| Neither model has a result on the other 34 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 11.1 | 9.0 | GLM 4.5 Air |
| Epoch Capabilities Index | Not measured | 134.0 | Only Mistral Medium 3 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 46.5% | 24.3% | GLM 4.5 Air |
| Neither model has a result on the other 40 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index50
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 73.3% | 57.8% | GLM 4.5 Air |
| Humanity's Last Exam (Artificial Analysis run) | 7.0% | 4.1% | GLM 4.5 Air |
| LMArena Text: hard prompts | 1379 | 1364 | GLM 4.5 Air |
| LMArena Text: expert | 1370 | 1342 | GLM 4.5 Air |
| LMArena Text: hard prompts in English | 1388 | 1374 | GLM 4.5 Air |
| LMArena Vision: diagrams | Not measured | 1155 | Only Mistral Medium 3 measured |
| LMArena Vision: homework | Not measured | 1165 | Only Mistral Medium 3 measured |
| Kagi LLM Benchmark | 43.0% | Not measured | Only GLM 4.5 Air measured |
| Neither model has a result on the other 14 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index40
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| AA Math Index | 80.7 | 30.3 | GLM 4.5 Air |
| AIME 2025 (Artificial Analysis run) | 80.7% | 30.3% | GLM 4.5 Air |
| LMArena Text: math | 1396 | 1353 | GLM 4.5 Air |
| LMArena Text: mathematical work | 1396 | 1354 | GLM 4.5 Air |
| Neither model has a result on the other 9 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index41
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 81.5% | 76.0% | GLM 4.5 Air |
| LMArena Text: science | 1407 | 1381 | GLM 4.5 Air |
| LMArena Text: medicine and healthcare | 1402 | 1381 | GLM 4.5 Air |
| LMArena Text: legal and government | 1382 | 1384 | Mistral Medium 3 |
| LMArena Text: business and finance | 1371 | 1352 | GLM 4.5 Air |
| Neither model has a result on the other 13 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index60
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| Design Arena: websites | 1146 | 1079 | GLM 4.5 Air |
| Design Arena: UI components | 1120 | 1022 | GLM 4.5 Air |
| Design Arena: game dev | 1094 | 1017 | GLM 4.5 Air |
| Design Arena: data viz | 1186 | 1034 | GLM 4.5 Air |
| Design Arena: 3D | 1122 | 1083 | GLM 4.5 Air |
| Design Arena: all code categories | 1137 | 1068 | GLM 4.5 Air |
| Design Arena: SVG | 1071 | Not measured | Only GLM 4.5 Air measured |
| Neither model has a result on the other 2 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index154
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | Leader |
|---|---|---|---|
| IFBench | 37.6% | 39.3% | Mistral Medium 3 |
| Long context reasoning (AA-LCR) | 46.7% | 31.3% | GLM 4.5 Air |
| DTBench | Not measured | 62.3% | Only Mistral Medium 3 measured |
| LMArena Text | 1384 | 1370 | GLM 4.5 Air |
| LMArena Text: instruction following | 1354 | 1338 | GLM 4.5 Air |
| LMArena Text: creative writing | 1343 | 1343 | Mistral Medium 3 |
| LMArena Vision | Not measured | 1157 | Only Mistral Medium 3 measured |
| LMArena Text: English | 1397 | 1385 | GLM 4.5 Air |
| LMArena Text: non-English | 1366 | 1353 | GLM 4.5 Air |
| LMArena Text: Chinese | 1426 | 1374 | GLM 4.5 Air |
| LMArena Text: German | 1377 | 1377 | GLM 4.5 Air |
| LMArena Text: Japanese | 1348 | 1316 | GLM 4.5 Air |
| LMArena Text: Korean | 1308 | 1304 | GLM 4.5 Air |
| LMArena Text: French | 1400 | 1368 | GLM 4.5 Air |
| LMArena Text: Spanish | 1385 | 1364 | GLM 4.5 Air |
| LMArena Text: Russian | 1373 | 1359 | GLM 4.5 Air |
| LMArena Text: Polish | 1368 | 1371 | Mistral Medium 3 |
| LMArena Text: multi-turn | 1371 | 1384 | Mistral Medium 3 |
| LMArena Text: longer queries | 1366 | 1360 | GLM 4.5 Air |
| LMArena Text: writing and language | 1354 | 1352 | GLM 4.5 Air |
| LMArena Text: entertainment and media | 1339 | 1335 | GLM 4.5 Air |
| LMArena Vision: English | Not measured | 1172 | Only Mistral Medium 3 measured |
| LMArena Vision: Chinese | Not measured | 1161 | Only Mistral Medium 3 measured |
| LMArena Vision: reading text | Not measured | 1154 | Only Mistral Medium 3 measured |
| LMArena Vision: humor | Not measured | 1145 | Only Mistral Medium 3 measured |
| LMArena Vision: captioning | Not measured | 1094 | Only Mistral Medium 3 measured |
| LMArena Vision: recognizing things | Not measured | 1141 | Only Mistral Medium 3 measured |
| LMArena Vision: creative writing | Not measured | 1146 | Only Mistral Medium 3 measured |
| Neither model has a result on the other 18 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | GLM 4.5 Air | Mistral Medium 3 | 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 | Z.ai GLM 4.5 Air | Mistral Medium 3 |
|---|---|---|
| Maker | Z.ai | Mistral AI |
| Weights | Open weights | No public weights listed |
| Released | July 25, 2025 | May 7, 2025 |
| Context window | 131K | 131K |
| Input $/Mtok | $0.20 | $0.40 |
| Output $/Mtok | $1.10 | $2.00 |
| 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 GLM 4.5 Air better than Mistral Medium 3 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick GLM 4.5 Air for output tokens 1.8x cheaper ($1.10 vs $2.00 per million) and work measured by LMArena Text: legal and government or IFBench, where it leads.
Which is cheaper, GLM 4.5 Air or Mistral Medium 3?
GLM 4.5 Air costs $0.20 per million input tokens and $1.10 per million output tokens; Mistral Medium 3 costs $0.40 and $2.00. Cost per solved task is known for neither of them.
Which has the bigger context window, GLM 4.5 Air or Mistral Medium 3?
Neither: both take 131K tokens.
How many benchmarks is the GLM 4.5 Air vs Mistral Medium 3 comparison based on?
45 benchmarks have a result for both models as of October 22, 2026. GLM 4.5 Air's coding index rests on 0 independent benchmarks and Mistral Medium 3's on 0. Where only one model has a result, the page says so instead of naming a winner.