---
license: apache-2.0
base_model: internlm/internlm3-8b-instruct
tags:
- gguf
- quantized
- internlm
- text-generation
- long-context
language:
- en
- zh
pipeline_tag: text-generation
---

# InternLM3-8B-Instruct — GGUF Quants

Quantized GGUF versions of [internlm/internlm3-8b-instruct](https://huggingface.co/internlm/internlm3-8b-instruct) — Shanghai AI Lab's InternLM3 8B instruction-tuned model featuring a **1M token context window**, strong multilingual support (English + Chinese), and competitive performance across reasoning and coding benchmarks.

The 1M context window makes InternLM3-8B uniquely capable among sub-10B models for long-document tasks, RAG pipelines, and extended reasoning chains.

## Available Files

| File | Quant | Size | Use Case |
|------|-------|------|----------|
| `InternLM3-8B-Instruct-Q8_0.gguf` | Q8_0 | ~8.5GB | Maximum quality |
| `InternLM3-8B-Instruct-Q6_K.gguf` | Q6_K | ~6.6GB | Near-lossless |
| `InternLM3-8B-Instruct-Q5_K_M.gguf` | Q5_K_M | ~5.7GB | High quality |
| `InternLM3-8B-Instruct-Q4_K_M.gguf` | Q4_K_M | ~4.9GB | **Recommended default** |
| `InternLM3-8B-Instruct-Q3_K_M.gguf` | Q3_K_M | ~3.9GB | Low VRAM |
| `InternLM3-8B-Instruct-IQ4_XS.gguf` | IQ4_XS | ~4.3GB | Imatrix 4-bit |
| `InternLM3-8B-Instruct-IQ3_XXS.gguf` | IQ3_XXS | ~3.2GB | Imatrix 3-bit |
| `InternLM3-8B-Instruct-IQ2_M.gguf` | IQ2_M | ~2.8GB | Imatrix 2-bit |
| `InternLM3-8B-Instruct-IQ1_S.gguf` | IQ1_S | ~2.0GB | Extreme compression |
| `InternLM3-8B-Instruct-fp16.gguf` | FP16 | ~16.0GB | Full precision |
| `imatrix.dat` | — | — | Importance matrix |

## Usage

```bash
# llama.cpp
./llama-cli -m InternLM3-8B-Instruct-Q4_K_M.gguf \
  --ctx-size 8192 -n 512 \
  -p "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n"

# Ollama
ollama run hf.co/DuoNeural/InternLM3-8B-Instruct-GGUF:Q4_K_M
```

## About InternLM3-8B

- **Parameters**: 8B
- **Context**: 1M tokens (unique at this parameter scale)
- **Architecture**: Decoder-only transformer
- **Languages**: English, Chinese (multilingual)
- **Strengths**: Long-context reasoning, instruction following, coding, math

Notable for its extreme context length — 1M tokens in a sub-10B model is unmatched in the open-source landscape.

---

*Quantized by DuoNeural using llama.cpp on RTX 5090.*

---

## DuoNeural

**DuoNeural** is an open AI research lab — human + AI in collaboration.

| Platform | Link |
|----------|------|
| HuggingFace | [huggingface.co/DuoNeural](https://huggingface.co/DuoNeural) |
| Website | [duoneural.com](https://duoneural.com) |
| GitHub | [github.com/DuoNeural](https://github.com/DuoNeural) |
| X / Twitter | [@DuoNeural](https://x.com/DuoNeural) |
| Email | duoneural@proton.me |
| Newsletter | [duoneural.beehiiv.com](https://duoneural.beehiiv.com) |
| Support | [buymeacoffee.com/duoneural](https://buymeacoffee.com/duoneural) |

### DuoNeural Research Publications

| Title | DOI |
|-------|-----|
| [Nano-CTM: Ternary Continuous Thought Machines with Thought-Space Self-Prediction for Efficient Iterative Reasoning](https://doi.org/10.5281/zenodo.19775622) | [10.5281/zenodo.19775622](https://doi.org/10.5281/zenodo.19775622) |
| [Recurrence as World Model: CTM Learns Implicit Belief States in Partially Observable Physical Environments](https://doi.org/10.5281/zenodo.19810620) | [10.5281/zenodo.19810620](https://doi.org/10.5281/zenodo.19810620) |
| [Per-Object Slot Decomposition for Scalable Neural World Modeling: When Does Attention Beat Mean-Field?](https://doi.org/10.5281/zenodo.19846804) | [10.5281/zenodo.19846804](https://doi.org/10.5281/zenodo.19846804) |
| [The Dynamical Horizon Principle: CTM Gates Converge to the Predictability Limit of Dynamical Systems](https://doi.org/10.5281/zenodo.19952612) | [10.5281/zenodo.19952612](https://doi.org/10.5281/zenodo.19952612) |

*Open access, CC BY 4.0. Authored by Archon, Jesse Caldwell, Aura — DuoNeural.*