---

license: mit

base_model:

- WeiboAI/VibeThinker-3B

library_name: gguf

pipeline_tag: text-generation

tags:

- gguf

- llama.cpp

- quantized

- unsloth

- lora

- trl

- sft

---

VibeThinker-3B-SFT-Claude-Opus-Reasoning-Unsloth-GGUF

GGUF quantizations of a LoRA fine-tune of WeiboAI/VibeThinker-3B, supervised fine-tuned on ermiaazarkhalili/claude-reasoning-distillation (private) (config sft).

Quantized from ermiaazarkhalili/VibeThinker-3B-SFT-Claude-Opus-Reasoning-Unsloth. See that repository for the full-precision weights.

| | |

| --- | --- |

| Base model | WeiboAI/VibeThinker-3B |

| Training data | ermiaazarkhalili/claude-reasoning-distillation (private) (config sft) |

| Method | LoRA supervised fine-tuning via Unsloth + TRL |

| License | mit (inherited from the base model) |

Available quantizations

| File | Size |

| --- | --- |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q2_k.gguf | 1.27 GB |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q3_k_m.gguf | 1.59 GB |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q4_k_m.gguf | 1.93 GB |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q5_k_m.gguf | 2.22 GB |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q6_k.gguf | 2.54 GB |

| vibethinker-3b-sft-claude-opus-reasoning-unsloth.q8_0.gguf | 3.29 GB |

Usage

llama.cpp

huggingface-cli download ermiaazarkhalili/VibeThinker-3B-SFT-Claude-Opus-Reasoning-Unsloth-GGUF vibethinker-3b-sft-claude-opus-reasoning-unsloth.q4_k_m.gguf --local-dir .
llama-cli -m vibethinker-3b-sft-claude-opus-reasoning-unsloth.q4_k_m.gguf -p "Explain gradient checkpointing in two sentences." -n 256

Ollama

echo 'FROM ./vibethinker-3b-sft-claude-opus-reasoning-unsloth.q4_k_m.gguf' > Modelfile
ollama create vibethinker-3b-sft-claude-opus-reasoning-unsloth-gguf -f Modelfile
ollama run vibethinker-3b-sft-claude-opus-reasoning-unsloth-gguf

Training configuration

| Setting | Value |

| --- | --- |

| LoRA rank (r) | 16 |

| LoRA alpha | 16 |

| Learning rate | 0.0002 |

| Epochs | 1 |

| Effective batch size | 8 (2 x 4 grad accum) |

| Max sequence length | 2048 |

| Base precision | 4-bit (QLoRA) |

| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |

Observed training loss

Measured from our SLURM logs for this configuration. These are training-loss

observations only — no downstream benchmark evaluation has been run on this

model, so they should not be read as a quality claim.

| SLURM job | Steps | First loss | Final loss |

| --- | --- | --- | --- |

| 45169145 | 1,310 | 2.1056 | 1.2599 |

Limitations

• No benchmark evaluation has been run on this checkpoint. The only reported

numbers are training-loss observations.

• Inherits the biases, knowledge cutoff and failure modes of the base model.
• Fine-tuned on a single instruction-following dataset; behaviour outside that

distribution is untested.

• LoRA adapters were merged into the base weights, so the merged model cannot

be detached from this fine-tune.

Reproducing

Trained by notebooks/sft_distillation_vibethinker-3b_unsloth.ipynb, executed non-interactively with

papermill on a SLURM H100 partition (Unsloth + TRL, LoRA).

---

Card generated from the training run's own configuration and logs by

scripts/generate_hub_model_card.py.