Model comparison
GLM-4.6 vs GLM-5.3
GLM-5.3 is the stronger model overall, scoring 54.8 to 41.4 on the Noometry Index. GLM-4.6 costs 2.1× less per token, which makes it the better buy when GLM-5.3's lead doesn't matter for your workload.
Last verified . 22 shared benchmarks.
Summary
- They share 22 benchmarks with published results for both. GLM-4.6 scores higher in 0 categories and GLM-5.3 in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GLM-5.3 leads 62.3 to 39.1.
- The biggest single-benchmark swing is SciCode: 38.4% for GLM-4.6 and 59% for GLM-5.3.
- GLM-4.6 is cheaper at $0.60 / $2.20 per million input/output tokens, against $1.40 / $4.40 for GLM-5.3.
- GLM-5.3 accepts more context: 1M tokens versus 205K.
Side by side
| GLM-4.6 | GLM-5.3 | |
|---|---|---|
| Provider | Z.ai (Zhipu) | Z.ai (Zhipu) |
| Noometry Index | 41.4 | 54.8 |
| Released | 2025-09-30 | 2026-08-14 |
| Weights | Open | Open |
| Context window | 205K | 1M |
| Max output | 131K | 131K |
| Input $ / M tokens | $0.60 | $1.40 |
| Output $ / M tokens | $2.20 | $4.40 |
| Results tracked | 29 | 42 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GLM-5.3 leads
GLM-4.6: 40.1 (#148), GLM-5.3: 59.5 (#14)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena WebDev | 1340 | 1622 |
| SciCode | 38.4% | 59% |
| LMArena Coding | 1449 | 1496 |
| ALE-Bench | 340.82 | 1,317 |
| DeepSWE | — | 69% |
| FrontierCode | — | 40.1% |
| SWE-bench Verified (bash only) | 55.4% | — |
| CursorBench | — | 42.6% |
| FrontierSWE | — | 30.2% |
| WeirdML | — | 75.4% |
Agentic & Tool Use GLM-5.3 leads
GLM-4.6: 32.3 (#66), GLM-5.3: 36.4 (#38)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| Terminal-Bench | 24.5% | — |
| APEX-Agents | — | 56.6% |
| Berkeley Function Calling Leaderboard | 72.4% | — |
| Vending-Bench 2 | — | 8,164 |
Reasoning GLM-5.3 leads
GLM-4.6: 23.7 (#172), GLM-5.3: 46.1 (#46)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| CritPt | 1.1% | 19.1% |
| LMArena Hard Prompts | 1440 | 1489 |
| Kagi LLM Benchmark | 47.4% | — |
| NYT Connections (extended) | — | 74.2% |
| Chess Puzzles | — | 21% |
| Mystery Game Puzzles | — | 33% |
| DTBench | — | 87.7% |
| LMCA | — | 55.5% |
| Bench to the Future 3 | — | 0.15 |
| Epoch Capabilities Index | — | 155.61 |
Math GLM-5.3 leads
GLM-4.6: 39.1 (#111), GLM-5.3: 62.3 (#33)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Math | 1432 | 1489 |
| FrontierMath (Tiers 1-3) | — | 68.8% |
| FrontierMath Tier 4 | — | 29.3% |
| OTIS Mock AIME 2024-2025 | — | 91.1% |
| ProofBench | — | 49% |
| FrontierMath (Feb 2025 set) | 3.8% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GLM-5.3 leads
GLM-4.6: 40.2 (#124), GLM-5.3: 58.3 (#37)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Expert | 1431 | 1516 |
| GPQA Diamond | — | 90.9% |
| SimpleQA Verified | — | 41% |
| Vectara Hallucination Rate | 9.5% | — |
Multilingual GLM-5.3 leads
GLM-4.6: 53.5 (#66), GLM-5.3: 55.7 (#28)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Non-English | 1426 | 1457 |
| LMArena Chinese | 1499 | 1528 |
| LMArena French | 1459 | 1499 |
| LMArena German | 1447 | 1499 |
| LMArena Japanese | 1393 | 1453 |
| LMArena Korean | 1400 | 1472 |
| LMArena Russian | 1419 | 1463 |
| LMArena Spanish | 1436 | 1460 |
Instruction Following GLM-5.3 leads
GLM-4.6: 74.3 (#98), GLM-5.3: 77.5 (#23)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Instruction Following | 1410 | 1477 |
Long Context GLM-5.3 leads
GLM-4.6: 43.4 (#94), GLM-5.3: 45.4 (#41)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Longer Query | 1422 | 1482 |
Writing & Preference GLM-5.3 leads
GLM-4.6: 61.1 (#90), GLM-5.3: 75.7 (#6)
| Benchmark | GLM-4.6 | GLM-5.3 |
|---|---|---|
| LMArena Text | 1440 | 1471 |
| LMArena Creative Writing | 1411 | 1457 |
| EQ-Bench Creative Writing | 1411 | 2075 |
| LMArena Multi-Turn | 1427 | 1472 |
Frequently asked questions
Is GLM-4.6 better than GLM-5.3?
GLM-5.3 is the stronger model overall, scoring 54.8 to 41.4 on the Noometry Index. GLM-4.6 costs 2.1× less per token, which makes it the better buy when GLM-5.3's lead doesn't matter for your workload.
Which is cheaper, GLM-4.6 or GLM-5.3?
GLM-4.6 is cheaper. It lists at $0.60 per million input tokens and $2.20 per million output tokens; GLM-5.3 lists at $1.40 and $4.40.
Is GLM-4.6 or GLM-5.3 better for coding?
GLM-5.3 scores higher on coding benchmarks: 59.5 versus 40.1 in the Noometry coding category.
Which has the bigger context window?
GLM-5.3 does, with 1M tokens against 205K.
How many benchmarks do GLM-4.6 and GLM-5.3 share?
22 benchmarks have published results for both models. GLM-4.6 has 29 scored results on Noometry and GLM-5.3 has 42.