glm-5-3-flashx: capability breakdown
152.1 / 220
measured 12 of 22 rungs · effort row: max (the median of its published effort settings)
Borrowed score: this is glm-5.3-flash's ladder, same weights on a faster stack. This endpoint has not been measured separately; every row below is glm-5.3-flash's.
This page shows where the number comes from. Each rung is one public benchmark, worth up to 10 points times the model's chance-corrected score on it; the score is the sum of all 22 rungs. How the capability ladder works explains the rungs, the fit and its limits.
| Rung (easiest → hardest) | Points / 10 | How | Source date |
|---|---|---|---|
| HumanEval | 9.96 | fitted | |
| MMLU | 9.81 | fitted | |
| MATH-500 | 9.93 | fitted | |
| MMLU-Pro | 9.59 | fitted | |
| AIME 2024 | 10.00 | fitted | |
| AIME 2025 | 9.91 | fitted | |
| LiveCodeBench | 9.53 | fitted | |
| GPQA Diamond | 8.83 | measured | 2026-09-05 |
| IFBench | 7.41 | fitted | |
| τ²-Telecom | 9.59 | fitted | |
| AA-LCR | 8.00 | measured | 2026-09-05 |
| Terminal-Bench 2.1 | 8.43 | measured | 2026-09-05 |
| SciCode | 5.16 | measured | 2026-09-05 |
| Terminal-Bench Hard | 6.20 | fitted | |
| AutomationBench | 6.04 | measured | 2026-09-13 |
| AA-Briefcase (rubric pass) | 4.64 | measured | 2026-09-05 |
| HLE | 3.68 | measured | 2026-09-05 |
| MLCR | 5.11 | measured | 2026-09-05 |
| τ³-Banking | 4.72 | measured | 2026-09-05 |
| Omniscience acc. | 2.75 | measured | 2026-09-05 |
| CritPt | 1.54 | measured | 2026-09-05 |
| GDP.pdf | 1.30 | measured | 2026-09-05 |
| Total | 152.13 | = 152.1 / 220 | |
Measured rows are the model's own published result on that benchmark. fitted rows are an item-response estimate: what a model with this ability would be expected to score on a benchmark it has no published result for (see the methodology).
Per-benchmark results are Artificial Analysis's (public model pages); fitted values are this site's estimates. Permission to republish requested 2026-09-05.