Model comparison
GLM-5 vs GPT-6 Astra
GPT-6 Astra is the stronger model overall, scoring 70.8 to 46.1 on the Noometry Index. GLM-5 costs 13× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Last verified . 30 shared benchmarks.
Summary
- They share 30 benchmarks with published results for both. GLM-5 scores higher in 2 categories and GPT-6 Astra in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Astra leads 85.1 to 27.6.
- The biggest single-benchmark swing is ARC-AGI-2: 4.9% for GLM-5 and 95% for GPT-6 Astra.
- GLM-5 is cheaper at $1 / $3.20 per million input/output tokens, against $10 / $50 for GPT-6 Astra.
- GPT-6 Astra accepts more context: 1.05M tokens versus 205K.
- GLM-5 has downloadable open weights; the other is API-only.
Side by side
| GLM-5 | GPT-6 Astra | |
|---|---|---|
| Provider | Z.ai (Zhipu) | OpenAI |
| Noometry Index | 46.1 | 70.8 |
| Released | 2026-02-11 | 2026-09-03 |
| Weights | Open | Proprietary |
| Context window | 205K | 1.05M |
| Max output | 131K | 128K |
| Input $ / M tokens | $1 | $10 |
| Output $ / M tokens | $3.20 | $50 |
| Results tracked | 45 | 56 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Astra leads
GLM-5: 49.0 (#52), GPT-6 Astra: 73.7 (#2)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena WebDev | 1434 | 1786 |
| WeirdML | 48.2% | 93.6% |
| LMArena Coding | 1461 | 1487 |
| ALE-Bench | 765.62 | 2,951 |
| SWE-bench Verified | 72.1% | — |
| DeepSWE | — | 74.1% |
| FrontierCode | — | 53.3% |
| SWE-bench Verified (bash only) | 72.8% | — |
| SWE-bench Multilingual | 69.7% | — |
| FrontierSWE | — | 65.5% |
| SciCode | — | 56.5% |
| GSO | — | 79.4% |
| MirrorCode | — | 46.7% |
Agentic & Tool Use GPT-6 Astra leads
GLM-5: 31.1 (#71), GPT-6 Astra: 52.9 (#3)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| Vending-Bench 2 | 4,432 | 15,515 |
| Terminal-Bench | 52.4% | — |
| APEX-Agents | — | 64.7% |
| Remote Labor Index | — | 20.8% |
| τ²-bench Airline | 82.5% | — |
| τ²-bench Banking | 9.8% | — |
| τ²-bench Retail | 73.7% | — |
| τ²-bench Telecom | 86.8% | — |
| BALROG | — | 68.3% |
| GDP.pdf | — | 34.2% |
Reasoning GPT-6 Astra leads
GLM-5: 27.6 (#116), GPT-6 Astra: 85.1 (#1)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| ARC-AGI-2 | 4.9% | 95% |
| NYT Connections (extended) | 74.8% | 98.1% |
| ARC-AGI-1 | 44.7% | 98.5% |
| Chess Puzzles | 10% | 72% |
| LMArena Hard Prompts | 1452 | 1462 |
| Epoch Capabilities Index | 145.83 | 166.45 |
| SimpleBench | 53.2% | — |
| Kagi LLM Benchmark | 75% | — |
| CritPt | — | 31.7% |
| EBR-Bench | — | 76.2% |
| Mystery Game Puzzles | — | 84% |
| DTBench | — | 97.3% |
| LMCA | — | 64.4% |
| Bench to the Future 3 | — | 0.14 |
| ForecastBench | 61 | — |
Math GPT-6 Astra leads
GLM-5: 46.4 (#71), GPT-6 Astra: 93.5 (#2)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 80% | 100% |
| LMArena Math | 1440 | 1465 |
| FrontierMath (Tiers 1-3) | — | 93.7% |
| FrontierMath Tier 4 | — | 97.6% |
| MathArena Final-Answer Competitions | 65.7% | — |
| ProofBench | — | 99% |
| FrontierMath (Feb 2025 set) | 16.4% | — |
| FrontierMath Erdős | — | 2.9% |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge GPT-6 Astra leads
GLM-5: 52.3 (#64), GPT-6 Astra: 75.3 (#1)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| GPQA Diamond | 87.8% | 95.8% |
| Vectara Hallucination Rate | 10.1% | 8.7% |
| LMArena Expert | 1454 | 1483 |
| Humanity's Last Exam | — | 54.8% |
| SimpleQA Verified | — | 75.6% |
Multimodal Not comparable
GLM-5: —, GPT-6 Astra: 55.0 (#3)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena Vision | — | 1281 |
| Blueprint-Bench 2 | — | 49.7% |
| Furniture Assembly | — | 80% |
| LMArena Document | — | 1468 |
Multilingual Too close to call
GLM-5: 53.7 (#58), GPT-6 Astra: 53.7 (#61)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena Non-English | 1430 | 1430 |
| LMArena Chinese | 1511 | 1484 |
| LMArena French | 1455 | 1456 |
| LMArena German | 1445 | 1440 |
| LMArena Japanese | 1416 | 1379 |
| LMArena Korean | 1423 | 1426 |
| LMArena Russian | 1436 | 1436 |
| LMArena Spanish | 1454 | 1407 |
Instruction Following GPT-6 Astra leads
GLM-5: 75.2 (#67), GPT-6 Astra: 76.3 (#44)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena Instruction Following | 1428 | 1450 |
Long Context Too close to call
GLM-5: 44.7 (#60), GPT-6 Astra: 44.5 (#62)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena Longer Query | 1446 | 1456 |
| CL-bench | 18.7% | — |
Writing & Preference GPT-6 Astra leads
GLM-5: 66.0 (#38), GPT-6 Astra: 75.3 (#7)
| Benchmark | GLM-5 | GPT-6 Astra |
|---|---|---|
| LMArena Text | 1446 | 1441 |
| LMArena Creative Writing | 1439 | 1418 |
| EQ-Bench Creative Writing | 1601 | 2173 |
| LMArena Multi-Turn | 1456 | 1448 |
Frequently asked questions
Is GLM-5 better than GPT-6 Astra?
GPT-6 Astra is the stronger model overall, scoring 70.8 to 46.1 on the Noometry Index. GLM-5 costs 13× less per token, which makes it the better buy when GPT-6 Astra's lead doesn't matter for your workload.
Which is cheaper, GLM-5 or GPT-6 Astra?
GLM-5 is cheaper. It lists at $1 per million input tokens and $3.20 per million output tokens; GPT-6 Astra lists at $10 and $50.
Is GLM-5 or GPT-6 Astra better for coding?
GPT-6 Astra scores higher on coding benchmarks: 73.7 versus 49.0 in the Noometry coding category.
Which has the bigger context window?
GPT-6 Astra does, with 1.05M tokens against 205K.
How many benchmarks do GLM-5 and GPT-6 Astra share?
30 benchmarks have published results for both models. GLM-5 has 45 scored results on Noometry and GPT-6 Astra has 56.