Model comparison
GLM-5 vs GPT-5.2
GPT-5.2 is the stronger model overall, scoring 54.1 to 46.1 on the Noometry Index. GLM-5 costs 3.1× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Last verified . 45 shared benchmarks.
Summary
- They share 45 benchmarks with published results for both. GLM-5 scores higher in 3 categories and GPT-5.2 in 6 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.2 leads 50.2 to 27.6.
- The biggest single-benchmark swing is ARC-AGI-2: 4.9% for GLM-5 and 52.9% for GPT-5.2.
- GLM-5 is cheaper at $1 / $3.20 per million input/output tokens, against $1.75 / $14 for GPT-5.2.
- GPT-5.2 accepts more context: 400K tokens versus 205K.
- GLM-5 has downloadable open weights; the other is API-only.
Side by side
| GLM-5 | GPT-5.2 | |
|---|---|---|
| Provider | Z.ai (Zhipu) | OpenAI |
| Noometry Index | 46.1 | 54.1 |
| Released | 2026-02-11 | 2025-12-11 |
| Weights | Open | Proprietary |
| Context window | 205K | 400K |
| Max output | 131K | 128K |
| Input $ / M tokens | $1 | $1.75 |
| Output $ / M tokens | $3.20 | $14 |
| Results tracked | 45 | 67 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.2 leads
GLM-5: 49.0 (#52), GPT-5.2: 51.6 (#37)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| SWE-bench Verified | 72.1% | 73.8% |
| SWE-bench Verified (bash only) | 72.8% | 72.8% |
| LMArena WebDev | 1434 | 1416 |
| SWE-bench Multilingual | 69.7% | 66.7% |
| WeirdML | 48.2% | 72.2% |
| LMArena Coding | 1461 | 1447 |
| ALE-Bench | 765.62 | 1,294 |
| GSO | — | 27.4% |
| AlgoTune | — | 2.05 |
Agentic & Tool Use GPT-5.2 leads
GLM-5: 31.1 (#71), GPT-5.2: 40.2 (#24)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| Terminal-Bench | 52.4% | 64.9% |
| τ²-bench Airline | 82.5% | 83% |
| τ²-bench Banking | 9.8% | 32.2% |
| τ²-bench Retail | 73.7% | 81.6% |
| τ²-bench Telecom | 86.8% | 89.7% |
| Vending-Bench 2 | 4,432 | 3,591 |
| Berkeley Function Calling Leaderboard | — | 55.9% |
| GDPval | — | 49.7% |
| Remote Labor Index | — | 2.5% |
| DeepResearch Bench | — | 41.1% |
| LMArena Search | — | 1207 |
| METR Time Horizons | — | 75.3% |
Reasoning GPT-5.2 leads
GLM-5: 27.6 (#116), GPT-5.2: 50.2 (#35)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| ARC-AGI-2 | 4.9% | 52.9% |
| SimpleBench | 53.2% | 45.8% |
| Kagi LLM Benchmark | 75% | 73.3% |
| NYT Connections (extended) | 74.8% | 83.6% |
| ARC-AGI-1 | 44.7% | 86.2% |
| Chess Puzzles | 10% | 49% |
| LMArena Hard Prompts | 1452 | 1445 |
| Epoch Capabilities Index | 145.83 | 153.45 |
| ForecastBench | 61 | 60.1 |
| EnigmaEval | — | 10.4% |
| EBR-Bench | — | 23% |
| Mystery Game Puzzles | — | 23% |
| DTBench | — | 90.9% |
| LMCA | — | 43.9% |
Math GPT-5.2 leads
GLM-5: 46.4 (#71), GPT-5.2: 60.0 (#38)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| MathArena Final-Answer Competitions | 65.7% | 72% |
| OTIS Mock AIME 2024-2025 | 80% | 96.1% |
| LMArena Math | 1440 | 1440 |
| FrontierMath (Feb 2025 set) | 16.4% | 40.7% |
| FrontierMath Tier 4 (v1) | 2.1% | 18.8% |
| FrontierMath (Tiers 1-3) | — | 67.4% |
| FrontierMath Tier 4 | — | 31.7% |
| ProofBench | — | 15% |
Knowledge GPT-5.2 leads
GLM-5: 52.3 (#64), GPT-5.2: 59.3 (#32)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| GPQA Diamond | 87.8% | 91.4% |
| Vectara Hallucination Rate | 10.1% | 8.4% |
| LMArena Expert | 1454 | 1445 |
| Humanity's Last Exam | — | 27.8% |
| SimpleQA Verified | — | 37.1% |
Multimodal Not comparable
GLM-5: —, GPT-5.2: 51.3 (#7)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| LMArena Vision | — | 1268 |
| VPCT | — | 84% |
| Furniture Assembly | — | 38.3% |
| LMArena Document | — | 1405 |
Multilingual Too close to call
GLM-5: 53.7 (#58), GPT-5.2: 53.4 (#67)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| LMArena Non-English | 1430 | 1425 |
| LMArena Chinese | 1511 | 1460 |
| LMArena French | 1455 | 1455 |
| LMArena German | 1445 | 1448 |
| LMArena Japanese | 1416 | 1420 |
| LMArena Korean | 1423 | 1392 |
| LMArena Russian | 1436 | 1440 |
| LMArena Spanish | 1454 | 1433 |
Instruction Following Too close to call
GLM-5: 75.2 (#67), GPT-5.2: 74.7 (#89)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| LMArena Instruction Following | 1428 | 1417 |
Long Context Too close to call
GLM-5: 44.7 (#60), GPT-5.2: 44.0 (#78)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| CL-bench | 18.7% | 18.2% |
| LMArena Longer Query | 1446 | 1428 |
Writing & Preference Too close to call
GLM-5: 66.0 (#38), GPT-5.2: 66.8 (#32)
| Benchmark | GLM-5 | GPT-5.2 |
|---|---|---|
| LMArena Text | 1446 | 1439 |
| LMArena Creative Writing | 1439 | 1401 |
| EQ-Bench Creative Writing | 1601 | 1703 |
| LMArena Multi-Turn | 1456 | 1458 |
Frequently asked questions
Is GLM-5 better than GPT-5.2?
GPT-5.2 is the stronger model overall, scoring 54.1 to 46.1 on the Noometry Index. GLM-5 costs 3.1× less per token, which makes it the better buy when GPT-5.2's lead doesn't matter for your workload.
Which is cheaper, GLM-5 or GPT-5.2?
GLM-5 is cheaper. It lists at $1 per million input tokens and $3.20 per million output tokens; GPT-5.2 lists at $1.75 and $14.
Is GLM-5 or GPT-5.2 better for coding?
GPT-5.2 scores higher on coding benchmarks: 51.6 versus 49.0 in the Noometry coding category.
Which has the bigger context window?
GPT-5.2 does, with 400K tokens against 205K.
How many benchmarks do GLM-5 and GPT-5.2 share?
45 benchmarks have published results for both models. GLM-5 has 45 scored results on Noometry and GPT-5.2 has 67.