GLM-5.3-Flash-Maya-GGUF
Maya-L: 99.2% of the full FP8 model's accuracy on zero-shot tasks (ARC, HellaSwag, WinoGrande, PIQA), in 156 GB. Maya-S: 97.9% in a 96 GB file. Maya-S24: 97.7%, and up to 14% faster decode on 24 GB cards, in 94.7 GB. Maya-M: 97.9%, and closer to the full model token by token, in 116 GB.
Params
321.3 B
Context
—
Downloads 30d
267 K
Likes
79
Download history
daily snapshots · 8 days267 K166 K
Oct 17Oct 19Oct 22Oct 24
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| mmproj-GLM-5.3-Flash-F16.gguf | GGUF | 1.1 GB | 49.9 GB | ❌ Won’t fit |
| GLM-5.3-Flash-Maya-S24-IQ2_XXS_S-00002-of-00003.gguf | IQ2_XXS_S | 47.0 GB | 100.4 GB | ❌ Won’t fit |
| GLM-5.3-Flash-Maya-M-IQ2_S-00002-of-00003.gguf | IQ2_S | 47.8 GB | 101.3 GB | ❌ Won’t fit |
| GLM-5.3-Flash-Maya-S-v2-IQ2_XXS-00001-of-00003.gguf | IQ2_XXS | 47.8 GB | 101.3 GB | ❌ Won’t fit |
| GLM-5.3-Flash-Maya-L-IQ3_S-00002-of-00004.gguf | IQ3_S | 48.0 GB | 101.5 GB | ❌ Won’t fit |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ ollama run glm-5-3-flash-maya-gguf # pin the quantization explicitly $ ollama run glm-5-3-flash-maya-gguf-iq3_s
$ huggingface-cli download peasantsmith/GLM-5.3-Flash-Maya-GGUF-GGUF \
GLM-5.3-Flash-Maya-L-IQ3_S-00002-of-00004.gguf --local-dir .
$ llama-cli -m GLM-5.3-Flash-Maya-L-IQ3_S-00002-of-00004.gguf \
-c 8192 -ngl 99 -t 8 --color
$ curl -s https://aimodelscomparison.com/api/v1/models/glm-5-3-flash-maya-gguf
{
"hf_id": "peasantsmith/GLM-5.3-Flash-Maya-GGUF",
"params_b": 321.32,
"context_length": null,
"license": { "id": "mit", "commercial": "yes" },
"downloads_30d": 267385,
"vram_estimates": [
{ "quant": "IQ3_S", "gb": 101.5 },
{ "quant": "IQ2_S", "gb": 101.3 }
],
"updated_at": "2026-10-17T01:00:19Z"
}
Specifications
- Parameters
- 321.3 B
- Licence
- mit
- First seen on the Hub
- 2026-10-07
- Base model
- GLM-5.3-Flash
- Training datasets
- undisclosed
- Added to our catalog
- 2026-10-17
Family
Base model and the most-downloaded derivatives in the catalog.
Compare with any text-generation model
Popular comparisons
Appears in