Model comparison
Gemma 3 4B vs GLM-5.2
GLM-5.2 is the stronger model overall, scoring 51.1 to 28.1 on the Noometry Index. Gemma 3 4B costs 43× less per token, which makes it the better buy when GLM-5.2's lead doesn't matter for your workload.
Last verified . 20 shared benchmarks.
GoogleSummary
- They share 20 benchmarks with published results for both. Gemma 3 4B scores higher in 0 categories and GLM-5.2 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where GLM-5.2 leads 57.1 to 11.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 7.5% for Gemma 3 4B and 86.4% for GLM-5.2.
- Gemma 3 4B is cheaper at $0.04 / $0.08 per million input/output tokens, against $1.40 / $4.40 for GLM-5.2.
- GLM-5.2 accepts more context: 1M tokens versus 131K.
Side by side
| Gemma 3 4B | GLM-5.2 | |
|---|---|---|
| Provider | Z.ai (Zhipu) | |
| Noometry Index | 28.1 | 51.1 |
| Released | 2025-03-12 | 2026-06-13 |
| Weights | Open | Open |
| Context window | 131K | 1M |
| Max output | 4K | 131K |
| Input $ / M tokens | $0.04 | $1.40 |
| Output $ / M tokens | $0.08 | $4.40 |
| Results tracked | 22 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5.2 leads
Gemma 3 4B: 35.9 (#215), GLM-5.2: 51.3 (#41)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| LMArena Coding | 1230 | 1485 |
| SWE-bench Verified | — | 78.7% |
| DeepSWE | — | 43.8% |
| FrontierCode | — | 24.5% |
| LMArena WebDev | — | 1603 |
| SciCode | — | 50.5% |
| WeirdML | — | 70.1% |
| ALE-Bench | — | 1,047 |
Agentic & Tool Use GLM-5.2 leads
Gemma 3 4B: 20.9 (#142), GLM-5.2: 32.4 (#63)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| APEX-Agents | — | 45.2% |
| Berkeley Function Calling Leaderboard | 19.6% | — |
| τ²-bench Banking | — | 37.1% |
| PostTrainBench | — | 31.7% |
| GBAEval | — | 0% |
| Vending-Bench 2 | — | 8,314 |
Reasoning GLM-5.2 leads
Gemma 3 4B: 13.2 (#335), GLM-5.2: 42.3 (#52)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| Kagi LLM Benchmark | 25.2% | 62.6% |
| Chess Puzzles | 0% | 21% |
| LMArena Hard Prompts | 1253 | 1480 |
| DTBench | 50.9% | 93.6% |
| LMCA | 2.8% | 45.8% |
| Epoch Capabilities Index | 116.02 | 151.78 |
| ARC-AGI-2 | — | 22.8% |
| SimpleBench | — | 58.8% |
| NYT Connections (extended) | — | 74.3% |
| ARC-AGI-1 | — | 77% |
| CritPt | — | 20.9% |
| EBR-Bench | — | 9.5% |
| Mystery Game Puzzles | — | 19% |
| Surface Evolver Bench | — | 55.6% |
Math GLM-5.2 leads
Gemma 3 4B: 16.8 (#292), GLM-5.2: 55.7 (#43)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 7.5% | 86.4% |
| LMArena Math | 1239 | 1482 |
| FrontierMath (Tiers 1-3) | — | 59.2% |
| FrontierMath Tier 4 | — | 29.3% |
| MathArena Final-Answer Competitions | — | 67.6% |
| ProofBench | — | 35% |
Knowledge GLM-5.2 leads
Gemma 3 4B: 11.8 (#299), GLM-5.2: 57.1 (#40)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| GPQA Diamond | 23.2% | 91.9% |
| LMArena Expert | 1223 | 1486 |
| SimpleQA Verified | — | 34.2% |
| Vectara Hallucination Rate | 6.4% | — |
Multilingual GLM-5.2 leads
Gemma 3 4B: 42.5 (#194), GLM-5.2: 55.8 (#26)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| LMArena Non-English | 1273 | 1459 |
| LMArena German | 1281 | 1468 |
| LMArena Russian | 1294 | 1466 |
| LMArena Chinese | — | 1519 |
| LMArena French | — | 1479 |
| LMArena Japanese | — | 1451 |
| LMArena Korean | — | 1445 |
| LMArena Spanish | — | 1477 |
Instruction Following GLM-5.2 leads
Gemma 3 4B: 65.2 (#225), GLM-5.2: 76.9 (#34)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| LMArena Instruction Following | 1239 | 1465 |
Long Context GLM-5.2 leads
Gemma 3 4B: 38.7 (#194), GLM-5.2: 45.3 (#43)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| LMArena Longer Query | 1273 | 1479 |
Writing & Preference GLM-5.2 leads
Gemma 3 4B: 42.0 (#239), GLM-5.2: 70.4 (#21)
| Benchmark | Gemma 3 4B | GLM-5.2 |
|---|---|---|
| LMArena Text | 1291 | 1470 |
| LMArena Creative Writing | 1271 | 1462 |
| EQ-Bench Creative Writing | 1068 | 1757 |
| LMArena Multi-Turn | 1255 | 1469 |
| EQ-Bench 4 | — | 1222 |
Frequently asked questions
Is Gemma 3 4B better than GLM-5.2?
GLM-5.2 is the stronger model overall, scoring 51.1 to 28.1 on the Noometry Index. Gemma 3 4B costs 43× less per token, which makes it the better buy when GLM-5.2's lead doesn't matter for your workload.
Which is cheaper, Gemma 3 4B or GLM-5.2?
Gemma 3 4B is cheaper. It lists at $0.04 per million input tokens and $0.08 per million output tokens; GLM-5.2 lists at $1.40 and $4.40.
Is Gemma 3 4B or GLM-5.2 better for coding?
GLM-5.2 scores higher on coding benchmarks: 51.3 versus 35.9 in the Noometry coding category.
Which has the bigger context window?
GLM-5.2 does, with 1M tokens against 131K.
How many benchmarks do Gemma 3 4B and GLM-5.2 share?
20 benchmarks have published results for both models. Gemma 3 4B has 22 scored results on Noometry and GLM-5.2 has 51.