Model Overview
Model Features
Model Capabilities
Use Cases
🚀 Llamacpp imatrix Quantizations of Apriel-Nemotron-15b-Thinker by ServiceNow-AI
This project offers quantized versions of the Apriel-Nemotron-15b-Thinker model using llama.cpp. It provides various quantization types to balance between model quality and resource usage, and offers clear guidance on running and downloading the models.
🚀 Quick Start
- Run in LM Studio: You can easily run the quantized models in LM Studio.
- Run with llama.cpp: Directly run the models with llama.cpp, or any other llama.cpp based project.
✨ Features
- Multiple Quantization Types: Offers a wide range of quantization types, such as bf16, Q8_0, Q6_K_L, etc., to meet different needs for quality and resource usage.
- Online Repacking: Some quantization types support online repacking for better performance on ARM and AVX machines.
📦 Installation
Downloading using huggingface-cli
Click to view download instructions
First, make sure you have hugginface-cli installed:
pip install -U "huggingface_hub[cli]"
Then, you can target the specific file you want:
huggingface-cli download bartowski/ServiceNow-AI_Apriel-Nemotron-15b-Thinker-GGUF --include "ServiceNow-AI_Apriel-Nemotron-15b-Thinker-Q4_K_M.gguf" --local-dir ./
If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:
huggingface-cli download bartowski/ServiceNow-AI_Apriel-Nemotron-15b-Thinker-GGUF --include "ServiceNow-AI_Apriel-Nemotron-15b-Thinker-Q8_0/*" --local-dir ./
You can either specify a new local-dir (ServiceNow-AI_Apriel-Nemotron-15b-Thinker-Q8_0) or download them all in place (./)
💻 Usage Examples
Prompt format
<|system|>
You are a thoughtful and systematic AI assistant built by ServiceNow Language Models (SLAM) lab. Before providing an answer, analyze the problem carefully and present your reasoning step by step. After explaining your thought process, provide the final solution in the following format: [BEGIN FINAL RESPONSE] ... [END FINAL RESPONSE].
{system_prompt}
<|end|>
<|user|>
{prompt}
<|end|>
<|assistant|>
Here are my reasoning steps:
📚 Documentation
Download a file (not the whole branch)
Filename | Quant type | File Size | Split | Description |
---|---|---|---|---|
Apriel-Nemotron-15b-Thinker-bf16.gguf | bf16 | 29.96GB | false | Full BF16 weights. |
Apriel-Nemotron-15b-Thinker-Q8_0.gguf | Q8_0 | 15.92GB | false | Extremely high quality, generally unneeded but max available quant. |
Apriel-Nemotron-15b-Thinker-Q6_K_L.gguf | Q6_K_L | 12.62GB | false | Uses Q8_0 for embed and output weights. Very high quality, near perfect, recommended. |
Apriel-Nemotron-15b-Thinker-Q6_K.gguf | Q6_K | 12.29GB | false | Very high quality, near perfect, recommended. |
Apriel-Nemotron-15b-Thinker-Q5_K_L.gguf | Q5_K_L | 11.07GB | false | Uses Q8_0 for embed and output weights. High quality, recommended. |
Apriel-Nemotron-15b-Thinker-Q5_K_M.gguf | Q5_K_M | 10.65GB | false | High quality, recommended. |
Apriel-Nemotron-15b-Thinker-Q5_K_S.gguf | Q5_K_S | 10.39GB | false | High quality, recommended. |
Apriel-Nemotron-15b-Thinker-Q4_K_L.gguf | Q4_K_L | 9.61GB | false | Uses Q8_0 for embed and output weights. Good quality, recommended. |
Apriel-Nemotron-15b-Thinker-Q4_1.gguf | Q4_1 | 9.50GB | false | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. |
Apriel-Nemotron-15b-Thinker-Q4_K_M.gguf | Q4_K_M | 9.11GB | false | Good quality, default size for most use cases, recommended. |
Apriel-Nemotron-15b-Thinker-Q4_K_S.gguf | Q4_K_S | 8.66GB | false | Slightly lower quality with more space savings, recommended. |
Apriel-Nemotron-15b-Thinker-IQ4_NL.gguf | IQ4_NL | 8.64GB | false | Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference. |
Apriel-Nemotron-15b-Thinker-Q4_0.gguf | Q4_0 | 8.63GB | false | Legacy format, offers online repacking for ARM and AVX CPU inference. |
Apriel-Nemotron-15b-Thinker-Q3_K_XL.gguf | Q3_K_XL | 8.58GB | false | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. |
Apriel-Nemotron-15b-Thinker-IQ4_XS.gguf | IQ4_XS | 8.20GB | false | Decent quality, smaller than Q4_K_S with similar performance, recommended. |
Apriel-Nemotron-15b-Thinker-Q3_K_L.gguf | Q3_K_L | 7.99GB | false | Lower quality but usable, good for low RAM availability. |
Apriel-Nemotron-15b-Thinker-Q3_K_M.gguf | Q3_K_M | 7.40GB | false | Low quality. |
Apriel-Nemotron-15b-Thinker-IQ3_M.gguf | IQ3_M | 6.94GB | false | Medium-low quality, new method with decent performance comparable to Q3_K_M. |
Apriel-Nemotron-15b-Thinker-Q3_K_S.gguf | Q3_K_S | 6.71GB | false | Low quality, not recommended. |
Apriel-Nemotron-15b-Thinker-Q2_K_L.gguf | Q2_K_L | 6.45GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. |
Apriel-Nemotron-15b-Thinker-IQ3_XS.gguf | IQ3_XS | 6.42GB | false | Lower quality, new method with decent performance, slightly better than Q3_K_S. |
Apriel-Nemotron-15b-Thinker-IQ3_XXS.gguf | IQ3_XXS | 5.99GB | false | Lower quality, new method with decent performance, comparable to Q3 quants. |
Apriel-Nemotron-15b-Thinker-Q2_K.gguf | Q2_K | 5.79GB | false | Very low quality but surprisingly usable. |
Apriel-Nemotron-15b-Thinker-IQ2_M.gguf | IQ2_M | 5.35GB | false | Relatively low quality, uses SOTA techniques to be surprisingly usable. |
Apriel-Nemotron-15b-Thinker-IQ2_S.gguf | IQ2_S | 4.98GB | false | Low quality, uses SOTA techniques to be usable. |
Apriel-Nemotron-15b-Thinker-IQ2_XS.gguf | IQ2_XS | 4.72GB | false | Low quality, uses SOTA techniques to be usable. |
Embed/output weights
Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.
🔧 Technical Details
ARM/AVX information
Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.
Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.
As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.
Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.
Click to view Q4_0_X_X information (deprecated)
I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking.
Click to view benchmarks on an AVX2 system (EPYC7702)
model | size | params | backend | threads | test | t/s | % (vs Q4_0) |
---|---|---|---|---|---|---|---|
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp512 | 204.03 ± 1.03 | 100% |
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp1024 | 282.92 ± 0.19 | 100% |
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | pp2048 | 259.49 ± 0.44 | 100% |
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg128 | 39.12 ± 0.27 | 100% |
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg256 | 39.31 ± 0.69 | 100% |
qwen2 3B Q4_0 | 1.70 GiB | 3.09 B | CPU | 64 | tg512 | 40.52 ± 0.03 | 100% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp512 | 301.02 ± 1.74 | 147% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp1024 | 287.23 ± 0.20 | 101% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | pp2048 | 262.77 ± 1.81 | 101% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg128 | 18.80 ± 0.99 | 48% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg256 | 24.46 ± 3.04 | 83% |
qwen2 3B Q4_K_M | 1.79 GiB | 3.09 B | CPU | 64 | tg512 | 36.32 ± 3.59 | 90% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | pp512 | 204.03 ± 1.03 | 100% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | pp1024 | 282.92 ± 0.19 | 100% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | pp2048 | 259.49 ± 0.44 | 100% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | tg128 | 39.12 ± 0.27 | 100% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | tg256 | 39.31 ± 0.69 | 100% |
qwen2 3B Q4_0_8_8 | 1.70 GiB | 3.09 B | CPU | 64 | tg512 | 40.52 ± 0.03 | 100% |
📄 License
This project is licensed under the MIT license.
Property | Details |
---|---|
Quantized by | bartowski |
Pipeline tag | text-generation |
License | mit |
Base model relation | quantized |
Base model | ServiceNow-AI/Apriel-Nemotron-15b-Thinker |

