EXAONE-3.5-7.8B-Instruct GGUF size and VRAM requirements
EXAONE-3.5-7.8B-Instruct-GGUF is a mid-size instruction-tuned chat model focused on general purpose & instruction. With approximately 7.82 billion parameters, it balances local hardware requirements with reasoning and generation fidelity. It is built on the exaone architecture with a native context window of up to 32,768 tokens. The model is primarily recommended for general knowledge, document summarization, email drafting, language translation, and multi-turn conversational chat. Released under the other license, it can be executed fully offline without sending data to external APIs.
Under the hood it uses 32 transformer layers, a hidden size of 4,096, 32 attention heads. It uses grouped-query attention (32 query heads sharing 8 key/value heads), which already trims KV-cache memory compared with full multi-head attention.
To run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF locally at a 4,096-token context, its quantized versions need between 3.76 GB (IQ2_S, lowest quality) and 30.43 GB (F32, 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 EXAONE-3.5-7.8B-Instruct-GGUF is about 0.5 GB, rising to roughly 4.0 GB at its full 32,768-token context. Shorter prompts free up memory for a higher-quality quantization.
For most users the best balance is Q6_K_L, needing about 7.47 GB. That means bartowski/EXAONE-3.5-7.8B-Instruct-GGUF fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.
Available GGUF quantizations for bartowski/EXAONE-3.5-7.8B-Instruct-GGUF include IQ2_S, IQ2_M, Q2_K, IQ3_XS, Q2_K_L, Q3_K_S, IQ3_M, Q3_K_M, Q3_K_L, IQ4_XS, Q4_0_4_4, Q4_0_4_8, Q4_0_8_8, Q4_0, IQ4_NL, Q4_K_S, Q3_K_XL, Q4_K_M, Q4_K_L, Q5_K_S, Q5_K_M, Q5_K_L, Q6_K, Q6_K_L, Q8_0, F16, F32. The model supports a native context length of up to 32,768 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_S | 2.7 | Low | 2.46 GB | 0.5 GB | 3.76 GB | 162.9 t/s | Fits in VRAM |
| IQ2_M | 2.89 | Low | 2.63 GB | 0.5 GB | 3.93 GB | 152.0 t/s | Fits in VRAM |
| Q2_K | 3.12 | Low | 2.84 GB | 0.5 GB | 4.14 GB | 140.6 t/s | Fits in VRAM |
| IQ3_XS | 3.46 | Fair | 3.15 GB | 0.5 GB | 4.45 GB | 127.0 t/s | Fits in VRAM |
| Q2_K_L | 3.54 | Fair | 3.23 GB | 0.5 GB | 4.53 GB | 124.0 t/s | Fits in VRAM |
| Q3_K_S | 3.61 | Fair | 3.29 GB | 0.5 GB | 4.59 GB | 121.7 t/s | Fits in VRAM |
| IQ3_M | 3.73 | Fair | 3.4 GB | 0.5 GB | 4.7 GB | 117.7 t/s | Fits in VRAM |
| Q3_K_M | 3.97 | Fair | 3.62 GB | 0.5 GB | 4.92 GB | 110.6 t/s | Fits in VRAM |
| Q3_K_L | 4.28 | Good | 3.9 GB | 0.5 GB | 5.2 GB | 102.6 t/s | Fits in VRAM |
| IQ4_XS | 4.4 | Good | 4.01 GB | 0.5 GB | 5.31 GB | 99.9 t/s | Fits in VRAM |
| Q4_0_4_4 | 4.62 | Good | 4.2 GB | 0.5 GB | 5.5 GB | 95.2 t/s | Fits in VRAM |
| Q4_0_4_8 | 4.62 | Good | 4.2 GB | 0.5 GB | 5.5 GB | 95.2 t/s | Fits in VRAM |
| Q4_0_8_8 | 4.62 | Good | 4.2 GB | 0.5 GB | 5.5 GB | 95.2 t/s | Fits in VRAM |
| Q4_0 | 4.63 | Good | 4.21 GB | 0.5 GB | 5.51 GB | 94.9 t/s | Fits in VRAM |
| IQ4_NL | 4.63 | Good | 4.22 GB | 0.5 GB | 5.52 GB | 94.9 t/s | Fits in VRAM |
| Q4_K_S | 4.65 | Good | 4.23 GB | 0.5 GB | 5.53 GB | 94.5 t/s | Fits in VRAM |
| Q3_K_XL | 4.66 | Good | 4.24 GB | 0.5 GB | 5.54 GB | 94.3 t/s | Fits in VRAM |
| Q4_K_M | 4.88 | Good | 4.44 GB | 0.5 GB | 5.74 GB | 90.0 t/s | Fits in VRAM |
| Q4_K_L | 5.2 | Very good | 4.73 GB | 0.5 GB | 6.03 GB | 84.5 t/s | Fits in VRAM |
| Q5_K_S | 5.56 | Very good | 5.06 GB | 0.5 GB | 6.36 GB | 79.0 t/s | Fits in VRAM |
| Q5_K_M | 5.7 | Very good | 5.19 GB | 0.5 GB | 6.49 GB | 77.1 t/s | Fits in VRAM |
| Q5_K_L | 5.96 | Very good | 5.43 GB | 0.5 GB | 6.73 GB | 73.7 t/s | Fits in VRAM |
| Q6_K | 6.57 | Excellent | 5.98 GB | 0.5 GB | 7.28 GB | 66.9 t/s | Fits in VRAM |
| Q6_K_L | 6.78 | Excellent | 6.17 GB | 0.5 GB | 7.47 GB | 64.9 t/s | Fits in VRAM |
| Q8_0 | 8.51 | Excellent | 7.74 GB | 0.5 GB | 9.04 GB | 6.5 t/s | Offload |
| F16 | 16.0 | Excellent | 14.57 GB | 0.5 GB | 15.87 GB | 3.4 t/s | Offload |
| F32 | 32.0 | Excellent | 29.13 GB | 0.5 GB | 30.43 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/EXAONE-3.5-7.8B-Instruct-GGUF?
bartowski/EXAONE-3.5-7.8B-Instruct-GGUF is an instruction-tuned chat model with 7.82 billion parameters, based on the exaone architecture. It is released under the other license and distributed as GGUF files for local inference.
How much VRAM do you need to run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF?
You need about 5.74 GB of VRAM to run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF entirely on the GPU using the Q4_K_M quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.
Can I run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF on an 8 GB GPU?
Yes. With 8 GB of VRAM you can run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF fully on the GPU using Q6_K_L (about 7.47 GB).
Can I run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF on a 16 GB GPU?
Yes. With 16 GB of VRAM you can run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF fully on the GPU using F16 (about 15.87 GB).
Can I run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF on a 24 GB GPU?
Yes. With 24 GB of VRAM you can run bartowski/EXAONE-3.5-7.8B-Instruct-GGUF fully on the GPU using F16 (about 15.87 GB).
What context length does bartowski/EXAONE-3.5-7.8B-Instruct-GGUF support?
bartowski/EXAONE-3.5-7.8B-Instruct-GGUF supports a native context length of up to 32,768 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/EXAONE-3.5-7.8B-Instruct-GGUF?
For bartowski/EXAONE-3.5-7.8B-Instruct-GGUF, a strong default is Q4_K_M, which needs about 5.74 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.