Up to four models
Claude Opus 5 vs GLM-5.3
On the index (GPT-4 = 100), Claude Opus 5 scores 185 and GLM-5.3 168. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Opus 5
185169–201
WeightsClosed
ReleasedJul 2026
Price in / out$5.0 / $25
Context1M
Core tests taken6 of 9
168152–184
WeightsOpen
ReleasedAug 2026
Price in / out$0.070 / $1.4
Context1.05M
Core tests taken5 of 9
Claude Opus 5 leads GLM-5.3 by 17 points, ahead on 5 of 5 shared core tests, at 25× the price.
Test by testthird-party results
Claude Opus 5
GLM-5.3
Every result as a table
| Test | Claude Opus 5 | GLM-5.3 |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.9% | 90.9% |
| LiveBench Data Analysis | 74.6% | 70.2% |
| LiveBench IF | 63.8% | 69.3% |
| LiveBench Language | 88.7% | 79.9% |
| LiveBench Reasoning | 91.2% | 85.8% |
| GPQA Diamond (Vals AI) | 93.4% | 88.1% |
| MMLU Pro | 91.6% | 86.8% |
| SimpleBench | 80.6% | 66.2% |
| SimpleQA Verified | 59.9% | 41.0% |
| Maths | ||
| FrontierMath (tiers 1-3) | 85.6% | 68.8% |
| AIME-style maths (OTIS mock) | 98.9% | 91.1% |
| LiveBench Mathematics | 95.7% | 87.9% |
| ProofBench v1.1 | 99.0% | 49.0% |
| Chess puzzles | 43.0% | 37.0% |
| FrontierMath (tier 4) | 73.2% | 29.3% |
| Coding | ||
| Terminal-Bench 4.0 | 53.5% | 38.9% |
| LiveBench Agentic Coding | 65.2% | 60.9% |
| LiveBench Coding | 81.5% | 79.0% |
| Code Migration | 57.5% | 44.2% |
| CyberBench v1.1 | 65.4% | 72.1% |
| IOI | 84.3% | 68.4% |
| LiveCodeBench | 89.0% | 80.5% |
| ProgramBench | 3.0% | 1.5% |
| SWE-bench | 97.0% | 95.4% |
| Vibe Code Bench 1-100 | 28.5% | 20.0% |
| Vibe Code Bench v1.1 | 88.4% | 78.1% |
| DeepSWE | 73.7% | 69.0% |
| FrontierSWE | 52.0% | 30.2% |
| WeirdML | 91.8% | 75.4% |
| Agents | ||
| APEX-Agents | 65.8% | 56.6% |
| SkillsBench | 60.4% | 47.5% |
| SRE Bench | 12.2% | 9.2% |
| Terminal-Bench-Science | 27.1% | 5.7% |
| DTBench | 97.9% | 87.7% |
| LMCA | 64.5% | 55.5% |
| Mystery game puzzles | 59.0% | 33.0% |
| Professional work | ||
| Finance Agent (v2) | 58.6% | 55.8% |
| EMB | 73.6% | 56.3% |
| LegalBench | 87.0% | 84.8% |
| Legal Research Bench | 55.3% | 49.0% |
| MedCode | 63.6% | 42.9% |
| MedScribe | 91.0% | 88.8% |
| Public Benefits Bench v1.1 | 76.9% | 68.5% |
| Tax Agent Bench | 47.1% | 43.3% |
| TaxEval v2 | 75.1% | 72.4% |
| Novel problems | ||
| MysteryMechanism | 37.4% | 23.0% |
Tests at least two of these models have taken. Bold rows are the core tests.