base_model, tags, license, language, datasets, pipeline_tag

base_model	tags	license	language	datasets	pipeline_tag
ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth	gguf llama.cpp unsloth qwen3 function-calling quantized	apache-2.0	en	Salesforce/xlam-function-calling-60k	text-generation

Qwen3-8B-xLAM-Unsloth — GGUF quantized

GGUF quantizations of ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth, produced via Unsloth + llama.cpp's conversion scripts.

Field	Value
Source checkpoint	ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth
Base model	unsloth/qwen3-8b-unsloth-bnb-4bit
Dataset	Salesforce/xlam-function-calling-60k
Training	1 full epoch (effective batch=8 via per_device=1 × grad_accum=8)
Final training loss	0.219 (job 36885898, runtime 3h 48m on H100 MIG 3g.40gb)
Conversion	Unsloth save_pretrained → llama.cpp convert_hf_to_gguf.py → llama-quantize
Quantization tool	llama.cpp llama-quantize (cached toolchain)

Available quantizations

File	Bits	Size	Notes
qwen3-8b-function-calling-xlam-unsloth.q2_k.gguf	2-bit	3.28 GB	Smallest; aggressive quality loss
qwen3-8b-function-calling-xlam-unsloth.q3_k_m.gguf	3-bit	4.12 GB	Small; noticeable quality loss
qwen3-8b-function-calling-xlam-unsloth.q4_k_m.gguf	4-bit	5.03 GB	Recommended — best size/quality balance
qwen3-8b-function-calling-xlam-unsloth.q5_k_m.gguf	5-bit	5.85 GB	Near-full quality
qwen3-8b-function-calling-xlam-unsloth.q6_k.gguf	6-bit	6.73 GB	Very close to Q8_0 at smaller size
qwen3-8b-function-calling-xlam-unsloth.q8_0.gguf	8-bit	8.71 GB	Largest; closest to bf16 source

Recommended default: Q4_K_M (4-bit, K-quant medium). For maximum fidelity, use Q8_0.

Usage

llama.cpp

# Text-only
llama-cli -hf ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth-GGUF --jinja -p "Find flights from SFO to NYC on December 25th" -n 256

# Interactive chat
llama-cli -hf ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth-GGUF --jinja -cnv

Ollama

ollama run hf.co/ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth-GGUF:Q4_K_M

llama-cpp-python

from llama_cpp import Llama
llm = Llama.from_pretrained(
    repo_id="ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth-GGUF",
    filename="*q4_k_m.gguf",
    n_ctx=2048,
)
out = llm.create_chat_completion(
    messages=[{"role": "user", "content": "Find flights from SFO to NYC on December 25th"}],
    max_tokens=256,
)
print(out["choices"][0]["message"]["content"])

Intended use

For research and non-commercial experimentation only. Outputs should be independently verified before any downstream use.

Limitations

• GGUF quantizations have unavoidable quality loss relative to the source bfloat16 checkpoint. Use Q5_K_M, Q6_K, or Q8_0 for best fidelity.
• Inherits all limitations of the source merged checkpoint (ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth).
• Limited to the function schemas covered in the xLAM-60K training dataset; performance on novel APIs may degrade.

Citation

@misc{ qwen3_8b_xlam_unsloth_2026_gguf ,
  author = {Ermia Azarkhalili},
  title = { Qwen3-8B-xLAM-Unsloth — GGUF quantized },
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/ermiaazarkhalili/Qwen3-8B-Function-Calling-xLAM-Unsloth-GGUF}}
}

---

This qwen3 model was trained 2× faster with Unsloth and Hugging Face's TRL library.