Z.ai (Zhipu AI) · zhipu/glm-5-3-flash
| Provider | Z.ai (Zhipu AI) |
|---|---|
| Family | glm |
| Type | vlm |
| Status | active |
| Released | 2026-08-26 |
| Updated | 2026-09-07 |
| Parameters | 321.3B |
| Open weights | yes |
Chain Of Thought · tier-2 Computer Use · tier-2 Function Calling · tier-2 Think Budget Control · tier-2
Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.
| Device | Memory | Bandwidth | Best quality | Weights | tok/s | Fastest |
|---|---|---|---|---|---|---|
| NVIDIA GB200 Grace Blackwell Superchip | 372.0 GB | 16000.0 GB/s | q6 | 240.99 GB | ~829.6 | ~1244.4 q4 |
| AMD Instinct MI355X | 288.0 GB | 8000.0 GB/s | q5 | 200.83 GB | ~497.8 | ~622.2 q4 |
| AMD Instinct MI325X | 256.0 GB | 6000.0 GB/s | q4 | 160.66 GB | ~466.7 | ~466.7 q4 |
| Apple M3 Ultra | 512.0 GB | 819.0 GB/s | fp8 | 321.32 GB | ~31.8 | ~63.7 q4 |
Taken from the card's availability section.
| Platform | Id on that platform |
|---|---|
| Fireworks AI | accounts/fireworks/models/glm-5p3-flash |
| Hugging Face | zai-org/GLM-5.3-Flash |
| Ollama | glm-5.3-flash |
| OpenRouter | z-ai/glm-5.3-flash |
| Together AI | zai-org/GLM-5.3-Flash |
| Zhipu GLM | glm-5.3-flash |
| Benchmark | Evaluated as | Score | Evidence date | Source kind | |
|---|---|---|---|---|---|
| aa_lcr | GLM-5.3-Flash | 80.0% | 2026-09-10 evaluated | independent evaluator | source |
| arena_elo_style_control | glm-5.3-flash | 1474.16 elo | 2026-09-10 evaluated | independent evaluator | source |
| critpt | GLM-5.3-Flash | 15.43% | 2026-09-10 evaluated | independent evaluator | source |
| gdpval_aa | GLM-5.3-Flash | 58.46% | 2026-09-10 evaluated | independent evaluator | source |
| gpqa_diamond | GLM-5.3-Flash | 91.21% | 2026-09-10 evaluated | independent evaluator | source |
| scicode | GLM-5.3-Flash | 51.62% | 2026-09-10 evaluated | independent evaluator | source |
This card reports no benchmark scores yet.
| Benchmark | Catalogue standing | Score |
|---|