Files and memory at 8K context
| Quant | File | Files | + KV + runtime | Smallest fit | Source |
|---|---|---|---|---|---|
| Q8_0 unsloth/GLM-5.3-Flash-GGUF | 318 GiB | 8 | unknown | over 192 GB GPU | HF files |
| BF16 unsloth/GLM-5.3-Flash-GGUF | 598 GiB | 14 | unknown | over 192 GB GPU | HF files |
| MLX 4-bit mixed precision mlx-community/GLM-5.3-Flash-4bit | 190 GiB | 62 | unknown | 256 GB Mac (cache unknown) | HF files |
| MLX 8-bit pipenetwork/GLM-5.3-Flash-MLX-8bit | 311 GiB | 32 | unknown | 512 GB Mac (cache unknown) | HF files |
No verified file: Q4_K_M, Q5_K_M, FP16. Sizes are exact file bytes (text weights; projectors listed separately). “+” = recurrent state or other terms unknown. Smallest fit adds 0.5 GiB spare; Macs use 75% of memory.
Quality
Artificial Analysis Intelligence Index
41.8
Variant: GLM-5.3-Flash · observed 2026-10-17 · source
Rank #5 of 75 researched 2026 models.
Arena text rating
1,471
95% CI 1466–1475 · 32,925 votes · published 2026-10-15 · dataset
Exact listed model name; hosted reasoning/quantization may differ from local checkpoint.
Scores describe the evaluated hosted variant, not each quantized file. AA and Arena are never averaged.
Reported speeds 5
| Hardware | Quant | Backend | Prompt t/s | Gen t/s | Date | Source |
|---|---|---|---|---|---|---|
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 4bit (GLM-5.3-Flash-oQ4e-mtp) | MLX (oMLX server) | 504 | 51.6 | 2026-10-16 | omlx.ai |
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 8bit (GLM-5.3-Flash-oQ8e-mtp) | MLX (oMLX server) | 508 | 64.8 | 2026-10-16 | omlx.ai |
| Apple M3 Ultra (80-core GPU), 512 GB unified memory | 4bit (GLM-5.3-Flash-oQ4e-mtp) | MLX (oMLX server) | 592 | 43.2 | 2026-10-16 | omlx.ai |
| Apple M5 Ultra (80-core GPU), 256 GB unified memory | 4bit (GLM-5.3-Flash-MLX-4bit) | MLX (oMLX server) | 2,402 | 139 | 2026-10-15 | omlx.ai |
| 1x NVIDIA B200 (datacenter GPU) | UD-IQ1_S | llama.cpp CUDA | 1,122 | 62.8 | 2026-09-30 | unsloth.ai |
Reported by the linked sources, not reproduced by LocalFit. All rows in the speed database