zai-org_GLM-4.7-Flash GGUF size and VRAM requirements
zai-org_GLM-4.7-Flash-GGUF is a very large language model focused on general purpose & instruction. With approximately 31.22 billion parameters, it balances local hardware requirements with reasoning and generation fidelity. It is built on the deepseek2 architecture with a native context window of up to 202,752 tokens. The model is primarily recommended for general knowledge, document summarization, email drafting, language translation, and multi-turn conversational chat. Released under the mit license, it can be executed fully offline without sending data to external APIs.
zai-org_GLM-4.7-Flash-GGUF is a Mixture-of-Experts model with 64 experts, of which 4 are active on each token. Routing only 4 of 64 experts makes it noticeably faster than a dense model of the same size — but every expert still has to be held in memory, so the numbers below are set by the full parameter count, not the active one. It is built from 47 transformer layers, a hidden size of 2,048, 20 attention heads. zai-org_GLM-4.7-Flash-GGUF uses Multi-head Latent Attention, caching a single compressed latent per layer instead of full keys and values. That keeps the KV cache small even at long context — the memory figures on this page already reflect it.
To run bartowski/zai-org_GLM-4.7-Flash-GGUF locally at a 4,096-token context, its quantized versions need between 8.11 GB (IQ2_XXS, lowest quality) and 56.8 GB (BF16, highest quality) of memory, weights plus KV cache and a system margin included. Context length drives that memory directly: at 4,096 tokens the KV cache for zai-org_GLM-4.7-Flash-GGUF is about 0.21 GB, rising to roughly 10.22 GB at its full 202,752-token context. Shorter prompts free up memory for a higher-quality quantization.
For most users the best balance is Q5_K_L, needing about 21.27 GB. That means bartowski/zai-org_GLM-4.7-Flash-GGUF fits entirely in the VRAM of a 10 GB GPU or larger, running fully on the GPU.
Available GGUF quantizations for bartowski/zai-org_GLM-4.7-Flash-GGUF include IQ2_XXS, IQ2_XS, IQ2_S, IQ2_M, Q2_K, Q2_K_L, IQ3_XXS, IQ3_XS, Q3_K_S, IQ3_M, Q3_K_M, Q3_K_L, Q3_K_XL, IQ4_XS, IQ4_NL, Q4_0, Q4_K_S, Q4_K_M, Q4_K_L, Q4_1, Q5_K_S, Q5_K_M, Q5_K_L, Q6_K, Q6_K_L, Q8_0, BF16. The model supports a native context length of up to 202,752 tokens; a longer context grows the KV cache and the memory needed.
Recommended Use Cases & Local Setup
Ideal for: General knowledge, document summarization, email drafting, language translation, and multi-turn conversational chat.
Simple one-line CLI installation running silently in the background
Polished desktop client with one-click model downloads and GPU offloading
Open-source privacy-focused desktop assistant
GGUF file size and memory by quantization
Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.
| Quant. | Bits | Quality | Weights | KV | Total | Speed~ | Verdict |
|---|---|---|---|---|---|---|---|
| IQ2_XXS | 1.95 | Very low | 7.1 GB | 0.21 GB | 8.11 GB | 7.0 t/s | Offload |
| IQ2_XS | 2.22 | Very low | 8.08 GB | 0.21 GB | 9.08 GB | 6.2 t/s | Offload |
| IQ2_S | 2.25 | Very low | 8.18 GB | 0.21 GB | 9.19 GB | 6.1 t/s | Offload |
| IQ2_M | 2.52 | Very low | 9.18 GB | 0.21 GB | 10.18 GB | 5.4 t/s | Offload |
| Q2_K | 2.83 | Low | 10.28 GB | 0.21 GB | 11.29 GB | 4.9 t/s | Offload |
| Q2_K_L | 2.91 | Low | 10.57 GB | 0.21 GB | 11.58 GB | 4.7 t/s | Offload |
| IQ3_XXS | 3.14 | Low | 11.41 GB | 0.21 GB | 12.42 GB | 4.4 t/s | Offload |
| IQ3_XS | 3.26 | Low | 11.84 GB | 0.21 GB | 12.84 GB | 4.2 t/s | Offload |
| Q3_K_S | 3.45 | Fair | 12.55 GB | 0.21 GB | 13.56 GB | 4.0 t/s | Offload |
| IQ3_M | 3.58 | Fair | 13.02 GB | 0.21 GB | 14.03 GB | 3.8 t/s | Offload |
| Q3_K_M | 3.61 | Fair | 13.11 GB | 0.21 GB | 14.11 GB | 3.8 t/s | Offload |
| Q3_K_L | 3.72 | Fair | 13.5 GB | 0.21 GB | 14.51 GB | 3.7 t/s | Offload |
| Q3_K_XL | 3.79 | Fair | 13.76 GB | 0.21 GB | 14.77 GB | 3.6 t/s | Offload |
| IQ4_XS | 4.16 | Fair | 15.13 GB | 0.21 GB | 16.14 GB | 3.3 t/s | Offload |
| IQ4_NL | 4.39 | Good | 15.97 GB | 0.21 GB | 16.98 GB | 3.1 t/s | Offload |
| Q4_0 | 4.46 | Good | 16.2 GB | 0.21 GB | 17.21 GB | 3.1 t/s | Offload |
| Q4_K_S | 4.55 | Good | 16.55 GB | 0.21 GB | 17.56 GB | 3.0 t/s | Offload |
| Q4_K_M | 4.73 | Good | 17.21 GB | 0.21 GB | 18.21 GB | 2.9 t/s | Offload |
| Q4_K_L | 4.79 | Good | 17.43 GB | 0.21 GB | 18.43 GB | 2.9 t/s | Offload |
| Q4_1 | 4.86 | Good | 17.65 GB | 0.21 GB | 18.66 GB | 2.8 t/s | Offload |
| Q5_K_S | 5.34 | Very good | 19.42 GB | 0.21 GB | 20.42 GB | 2.6 t/s | Offload |
| Q5_K_M | 5.53 | Very good | 20.09 GB | 0.21 GB | 21.09 GB | 2.5 t/s | Offload |
| Q5_K_L | 5.58 | Very good | 20.27 GB | 0.21 GB | 21.27 GB | 2.5 t/s | Offload |
| Q6_K | 6.36 | Very good | 23.12 GB | 0.21 GB | 24.13 GB | — | Insufficient |
| Q6_K_L | 6.4 | Very good | 23.26 GB | 0.21 GB | 24.27 GB | — | Insufficient |
| Q8_0 | 8.16 | Excellent | 29.66 GB | 0.21 GB | 30.66 GB | — | Insufficient |
| BF16 | 15.35 | Excellent | 55.79 GB | 0.21 GB | 56.8 GB | — | Insufficient |
KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.
Frequently asked questions
What kind of model is bartowski/zai-org_GLM-4.7-Flash-GGUF?
bartowski/zai-org_GLM-4.7-Flash-GGUF is a language model with 31.22 billion parameters, based on the deepseek2 architecture. It is released under the mit license and distributed as GGUF files for local inference.
Is bartowski/zai-org_GLM-4.7-Flash-GGUF a Mixture-of-Experts (MoE) model?
Yes. bartowski/zai-org_GLM-4.7-Flash-GGUF is a Mixture-of-Experts model with 64 experts, of which 4 are activated per token. That makes it faster than a dense model of the same size, but all 64 experts must be loaded into memory, so the VRAM/RAM it needs is driven by the total parameter count, not the active one.
How does bartowski/zai-org_GLM-4.7-Flash-GGUF handle long context?
bartowski/zai-org_GLM-4.7-Flash-GGUF uses Multi-head Latent Attention, which caches one compressed latent per layer instead of full keys and values. Its KV cache therefore grows much more slowly than a standard model of the same size, so long contexts add relatively little memory.
How much VRAM do you need to run bartowski/zai-org_GLM-4.7-Flash-GGUF?
You need about 9.19 GB of VRAM to run bartowski/zai-org_GLM-4.7-Flash-GGUF entirely on the GPU using the IQ2_S quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run bartowski/zai-org_GLM-4.7-Flash-GGUF on an 8 GB GPU?
Partially. bartowski/zai-org_GLM-4.7-Flash-GGUF only fits on an 8 GB GPU by offloading part of it to system RAM (with Q5_K_L), which runs but is slower.
Can I run bartowski/zai-org_GLM-4.7-Flash-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run bartowski/zai-org_GLM-4.7-Flash-GGUF fully on the GPU using Q3_K_XL (about 14.77 GB).
Can I run bartowski/zai-org_GLM-4.7-Flash-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run bartowski/zai-org_GLM-4.7-Flash-GGUF fully on the GPU using Q5_K_L (about 21.27 GB).
What context length does bartowski/zai-org_GLM-4.7-Flash-GGUF support?
bartowski/zai-org_GLM-4.7-Flash-GGUF supports a native context length of up to 202,752 tokens. A longer context grows the KV cache, so it increases the memory needed to run the model.
What is the best quantization for bartowski/zai-org_GLM-4.7-Flash-GGUF?
For bartowski/zai-org_GLM-4.7-Flash-GGUF, a strong default is Q4_K_M, which needs about 18.21 GB and keeps most of the quality while roughly halving the memory versus 8-bit. With VRAM to spare, Q5_K_M or Q6_K add a little more quality; if you are tight on memory, a smaller quantization still runs. Pick the highest quantization that fits your VRAM.