Views
No views yet
| Model | Size | GSM8K | IFEval | MMLU |
|---|---|---|---|---|
| qwen3-8b-lora-v2-Q3_K_M.gguf | 3.80 GB | 92.5% | 77.7% | 71.4% |
| qwen3-8b-lora-v3-Q3_K_M.gguf | 3.80 GB | 93.0% | 76.6% | 72.3% |
| qwen3-1.7b-distill-v3-Q8_0.gguf | 1.83 GB | 77.5% | 68.1% | 48.1% |
| qwen3-1.7b-distill-v3-Q3_K_M.gguf | 0.90 GB | 77.5% | 68.1% | 48.1% |
| qwen2.5-1.5b-distill-truncated-Q8_0.gguf | 1.50 GB | 75.0% | 47.9% | 63.2% |
| Need | Recommended |
|---|---|
| Best instruction following, ~4 GB | qwen3-8b-lora-v2 |
| Best math + knowledge, ~4 GB | qwen3-8b-lora-v3 |
| Compact, under 2 GB | qwen3-1.7b-distill-v3-Q8_0 |
| Ultra-compact, under 1 GB | qwen3-1.7b-distill-v3-Q3_K_M |
| Best knowledge breadth, under 2 GB | qwen2.5-1.5b-distill-truncated-Q8_0 |
1# Download Modelfile then create
2curl -L https://huggingface.co/tellerlin/evomerge-pareto-gguf/resolve/main/Modelfile-qwen3-8b-v2 -o Modelfile
3ollama create evomerge-qwen3-v2 -f Modelfile
4ollama run evomerge-qwen3-v2ollama run hf.co/tellerlin/evomerge-pareto-gguf:qwen3-8b-lora-v2-Q3_K_M⚠️ If you use a custom Modelfile: Do not include aTEMPLATEline. AddingTEMPLATE {{ .Prompt }}will override Qwen3's built-in chat template and break tool calling, thinking mode, and system prompts. Only setPARAMETERandSYSTEM— Ollama will automatically use the correct template from the GGUF file.
llama-cli -m qwen3-8b-lora-v2-Q3_K_M.gguf -p "Your prompt here" -n 512 -c 327681from llama_cpp import Llama
2llm = Llama("qwen3-8b-lora-v2-Q3_K_M.gguf", n_ctx=32768)
3out = llm("Your prompt", max_tokens=512)
4print(out["choices"][0]["text"])