GLM-5.3
| Released | 2026-08-14 |
|---|---|
| Price per 1M | $1.20 in · $4.18 out |
| Overall score | 60.1 · rank 29 |
| Context window | 1,000,000 tokens |
| Provider | Zhipu AI |
| Weights | Open weights |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 60.1 | 29 | 43 |
| Overall: Coding | 60.2 | 25 | 31 |
| Overall: Reasoning | 58.4 | 55 | 14 |
| Overall: Agentic work | 60.6 | 27 | 29 |
| Overall: Frontend | 59.1 | 31 | 12 |
| Coding | 60.6 | 24 | 10 |
| Agents | 60.8 | 21 | 12 |
| Reasoning | 59.3 | 52 | 7 |
| Math | 58.1 | 42 | 4 |
| Knowledge | 56.6 | 50 | 3 |
| Frontend | 58.4 | 36 | 2 |
| General | 64.7 | 10 | 5 |
| Value vs price | +10.3 | 28 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 61.4 ±4.2 | 5 |
| High | 59.3 ±3.5 | 3 |
| Low | 58.0 ±4.4 | 3 |
| Max | 60.0 ±0.9 | 36 |
Evidence
Coding 60.6 ±1.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| CursorBench | 42.6% | Epoch AI | 2026-11-11 |
| FrontierSWE | 30.2% | Epoch AI | 2026-11-11 |
| DeepSWE | 69.0% | Epoch AI | 2026-11-11 |
| FrontierCode | 40.1% | Epoch AI | 2026-11-11 |
| Code Migration v1.1 | 55.1% | Vals AI | 2026-11-09 |
| IOI | 68.4% | Vals AI | 2026-10-29 |
| SWE-bench Verified (Vals run) | 95.4% | Vals AI | 2026-09-01 |
| SciCode | 59.0% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 80.5% | Vals AI | 2026-09-01 |
| ProgramBench | 1.5% | Vals AI | 2026-11-09 |
Agents 60.8 ±0.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 4 | 41.2% | Terminal-Bench | 2026-08-14 |
| SRE Bench | 9.2% | Vals AI | 2026-11-05 |
| Terminal-Bench-Science | 5.7% | Vals AI | 2026-11-09 |
| APEX-Agents | 56.6% | Epoch AI | 2026-11-11 |
| Terminal-Bench 4.0 (Vals run) | 38.9% | Vals AI | 2026-11-09 |
| Terminal-Bench 2.1 (Vals run) | 71.5% | Vals AI | 2026-10-01 |
| EMB v1.1 (Excel financial modeling) | 69.1% | Vals AI | 2026-11-09 |
| Tax Agent Bench | 43.3% | Vals AI | 2026-11-09 |
| Finance Agent v2.1 | 65.0% | Vals AI | 2026-11-09 |
| SkillsBench | 47.5% | Vals AI | 2026-10-29 |
| Vending-Bench 2 | $8,164 | Epoch AI | 2026-11-11 |
| HLAB-Vals 1.1 (legal agent) | 14.6% | Vals AI | 2026-11-09 |
| CyberBench v1.1 | 72.1% | Vals AI | 2026-11-05 |
Reasoning 59.3 ±1.9
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Extended NYT Connections | 74.2% | Lech Mazur | 2026-11-10 |
| MysteryMechanism | 23.0% | Vals AI | 2026-11-09 |
| CritPt | 19.1% | Epoch AI | 2026-11-11 |
| SimpleBench | 66.2% | SimpleBench | 2026-08-19 |
| GPQA Diamond | 90.9% | Epoch AI | 2026-08-24 |
| GPQA Diamond (Vals run) | 88.1% | Vals AI | 2026-09-01 |
| Chess Puzzles | 37.0% | Epoch AI | 2026-10-13 |
| Mystery Game Puzzles | 33.0% | Epoch AI | 2026-08-30 |
Math 58.1 ±1.4
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierMath Tiers 1-3 | 68.8% | Epoch AI | 2026-08-25 |
| FrontierMath Tier 4 | 29.3% | Epoch AI | 2026-08-25 |
| OTIS Mock AIME 2024-2025 | 91.1% | Epoch AI | 2026-08-24 |
| ProofBench v1.1 | 49.0% | Vals AI | 2026-11-09 |
| ProofBench | 49.0% | Epoch AI | 2026-11-11 |
Knowledge 56.6 ±4.9
| Benchmark | Result | Source | Date |
|---|---|---|---|
| SimpleQA Verified | 41.0% | Epoch AI | 2026-08-28 |
| MMLU-Pro (Vals run) | 86.8% | Vals AI | 2026-09-01 |
| Legal Research Bench | 49.0% | Vals AI | 2026-11-09 |
Frontend 58.4 ±0.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Vibe Code Bench v1.1 | 78.1% | Vals AI | 2026-11-09 |
| LMArena WebDev | 1,622 | LMArena | 2026-11-09 |
General 64.7 ±2.9
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LLM Debate Benchmark | 1,634 | Lech Mazur | 2026-10-26 |
| EQ-Bench Creative Writing v3 | 2,076 | EQ-Bench | 2026-11-11 |
| LMArena Text | 1,483 | LMArena | 2026-11-09 |
| Creative Story-Writing (Lech Mazur) | 87.0% | Lech Mazur | 2026-10-23 |
| EQ-Bench Longform Creative Writing | 81.8% | EQ-Bench | 2026-11-11 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| CritPt | 14.6% | Epoch AI | A stronger configuration result was used |
| CursorBench | 38.0% | Epoch AI | A stronger configuration result was used |
| CursorBench | 33.3% | Epoch AI | A stronger configuration result was used |
| SciCode | 42.0% | Epoch AI | A stronger configuration result was used |
Names matched to this model
glm-5-3