Can the RTX 4070 run GLM-4.7-Flash?
Not fully, but it's usable. GLM-4.7-Flash needs about 19.8 GB at Q4_K_M and the RTX 4070 has 12 GB - yet as a mixture-of-experts model it reads only ~3B parameters per token, so you can keep attention on the GPU and the experts in system RAM (llama.cpp --n-cpu-moe) and still get interactive speeds with 32 GB+ of RAM.
Every quantization
GLM-4.7-Flash (31.2B, 3B active) on 12 GB
| Quant | Weights | Needs | Fit | Est. speed |
|---|---|---|---|---|
| Q4_K_M | 18 GB | 19.8 GB | Too big | - |
| Q5_K_M | 22 GB | 23.8 GB | Too big | - |
| Q8_0 | 33 GB | 34.8 GB | Too big | - |
| FP16 | 62 GB | 63.8 GB | Too big | - |
Estimates. Needs = weights + 1.8 GB for KV cache and runtime at modest context. Speed is single-stream decode scaled from the RTX 4070's 504 GB/s bandwidth (Estimated from memory bandwidth (llama.cpp decode)). Long contexts need more memory and run slower.
Run it with expert offload
Attention on the GPU, experts in system RAM
llama-server -hf unsloth/GLM-4.7-Flash-GGUF -ngl 99 --n-cpu-moe 99Lower --n-cpu-moe step by step until the RTX 4070's 12 GB is nearly full - each expert layer moved back to the GPU speeds things up. Needs about 16 GB+ of free system RAM.
What to do instead
Two ways forward
Cheapest cards that run it
- NVIDIA Tesla P40 - 24 GB, ~$159
- AMD Radeon RX 7900 XT - 20 GB, ~$699
- NVIDIA RTX 3090 - 24 GB, ~$750
- AMD Radeon RX 7900 XTX - 24 GB, ~$849
Models that fit the RTX 4070
- Qwen3 14B - 14B, 8 GB at Q4
- Gemma 4 12B - 12B, 7 GB at Q4
- Qwen3.5 9B - 9.7B, 6 GB at Q4
- Llama 3.1 8B - 8B, 5.5 GB at Q4
Compare
Same model, other hardware - and other models on this card
GLM-4.7-Flash on other hardware
- RTX 5090~396 tok/s
- RTX 5090 D~395 tok/s
- RTX 6000D~347 tok/s
- RTX 4090~241 tok/s
- Radeon RX 7900 XTX~212 tok/s
- RTX 3090~211 tok/s
- MacBook Pro 16" M4 Max~121 tok/s
- Arc Pro B60~101 tok/s
Also runs on the RTX 4070
- Qwen3 14B~36 tok/s
- Gemma 4 12B~41 tok/s
- Qwen3.5 9B~48 tok/s
- Llama 3.1 8B~53 tok/s
- Mistral-7B-Instruct-v0.2~72 tok/s
- Qwen3.5 4B~96 tok/s
- Qwen3-4B-Instruct-2507~145 tok/s
- NVIDIA Nemotron 3 Nano 4B~145 tok/s
Frequently asked
Quick answers
Can the NVIDIA RTX 4070 run GLM-4.7-Flash?
Not fully, but it's usable. GLM-4.7-Flash needs about 19.8 GB at Q4_K_M and the RTX 4070 has 12 GB - yet as a mixture-of-experts model it reads only ~3B parameters per token, so you can keep attention on the GPU and the experts in system RAM (llama.cpp --n-cpu-moe) and still get interactive speeds with 32 GB+ of RAM.
How much VRAM does GLM-4.7-Flash need?
About 18 GB of weights at Q4_K_M, 33 GB at Q8_0 and 62 GB at FP16. Add roughly 1.8 GB for the KV cache and runtime at modest context lengths.
How fast is GLM-4.7-Flash on the RTX 4070?
It doesn't fit in the RTX 4070's memory, so speed depends on how many layers you offload to system RAM - typically single-digit tokens per second or lower.