Llama-3.3-70B-Instruct-lecore-GGUF
GGUF builds of meta-llama/Llama-3.3-70B-Instruct for llama.cpp, Ollama and LM Studio. These are companions to the MLX export staccs/Llama-3.3-70B-Instruct-lecore-mlx, which has the measured perplexity numbers.
| File | Size | What it is |
|---|---|---|
Llama-3.3-70B-Instruct-Q4_K_M.gguf |
42.52 GB | stock llama.cpp Q4_K_M |
Llama-3.3-70B-Instruct-lecore-mix.gguf |
27.20 GB | leCore per-tensor mix: each linear layer gets the GGUF K-quant matching the bits leCore measured for it (2→Q2_K, 3→Q3_K, 4→Q4_K, 8→Q8_0); embeddings and output follow the MLX export |
The lecore-mix file is quantized from the original bf16 weights with llama.cpp K-quants. The bit width of each tensor follows leCore's measured (KL-damage) allocation, so its numbers are close to, but not identical to, the MLX export.
Some tensor widths are not multiples of 256, so llama.cpp may fall back to a compatible type for those tensors.
Usage
Ollama: download a file, then
printf 'FROM ./Llama-3.3-70B-Instruct-Q4_K_M.gguf
' > Modelfile && ollama create llama-3.3-70b-instruct-lecore-gguf -f Modelfile && ollama run llama-3.3-70b-instruct-lecore-gguf
or pull straight from the Hub with ollama run hf.co/staccs/Llama-3.3-70B-Instruct-lecore-GGUF:Q4_K_M
LM Studio: search for staccs/Llama-3.3-70B-Instruct-lecore-GGUF in the model browser and pick a file. On a 36 GB Mac, the lecore-mix file is the one that fits.
llama.cpp
llama-cli -hf staccs/Llama-3.3-70B-Instruct-lecore-GGUF -p "Hello" -n 256
Benchmarks (lm-eval-harness 0.4.9, run on the shipped weights dequantized to bf16 on GPU)
MMLU 5-shot on the first 10 questions of each of the 57 subjects; GSM8K 5-shot on the first 100 test problems (greedy, 320 new tokens, no chat template). These are subsets, so expect about ±2–4 points of sampling noise.
| Model | MMLU (5-shot) | GSM8K strict | GSM8K flexible |
|---|---|---|---|
| leCore mix (this repo) | 72.6% | 82.0% | 84.0% |
| stock 3-bit affine, group 64 (simulated) | 24.2% | 0.0% | 0.0% |
License
Base model: meta-llama/Llama-3.3-70B-Instruct (weights taken from the identical public mirror unsloth/Llama-3.3-70B-Instruct). It is released by Meta under the Llama 3.3 Community License Agreement (https://www.llama.com/llama3_3/license/) and its Acceptable Use Policy (https://www.llama.com/llama3_3/use-policy/). This derivative is distributed under the same license. Built with Llama. Llama 3.3 is licensed under the Llama 3.3 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved. Only quantization was done here: no retraining.
Related builds
Same base model, other formats in the lecore set:
| repo | format | size |
|---|---|---|
| Llama-3.3-70B-Instruct-lecore-hrr | gated-HRR gates | 5.4 MB |
| Llama-3.3-70B-Instruct-lecore-mlx | MLX (Apple Silicon) | 26.81 GB |