GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
saidutta69
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
saidutta69
Model Tree
HuggingFaceTB/SmolLM2-360M-Instruct
Input Modalities
Output Modalities
Supported Functionality
At this size there is no VRAM ladder to speak of — pick a quant by how much RAM and bandwidth your device has:
| Device | Recommended quant | Weights |
|---|---|---|
| Phone / browser (4 GB RAM) | Q4_K_M | ~0.25 GB |
| Raspberry Pi 5 (4 GB) | Q4_K_M | ~0.25 GB |
| Laptop (8 GB, Apple Silicon) | Q8_0 | ~0.36 GB |
| Server / multi-tenant | Q6_K | ~0.34 GB |
Weights only, at this model's 362M native size. The Q2_K tier (~0.20 GB) is there if you are truly RAM-bound, but at 362M the model is small enough that Q4_K_M already fits almost anywhere. Context is the real cost: add roughly 16 MB per 8K tokens of KV cache.
Trial 137 of a 200-trial Heretic run (seed 318129666).
| Parameter | Value |
|---|---|
| direction_index | 27.04 |
| attn.o_proj.max_weight | 1.32 |
| attn.o_proj.max_weight_position | 29.44 |
| attn.o_proj.min_weight | 0.46 |
| attn.o_proj.min_weight_distance | 16.37 |
| mlp.down_proj.max_weight | 1.19 |
| mlp.down_proj.max_weight_position | 30.27 |
| mlp.down_proj.min_weight | 0.22 |
| Metric | This model | Original model (HuggingFaceTB/SmolLM2-360M-Instruct) |
|---|---|---|
| KL divergence | 0.0154 | 0 (by definition) |
| Refusals | 4/100 | 8/100 |
Fine-tuning a "helpful" persona on top of RLHF'd refusals fights the base model's training and tends to degrade coherence. Abliteration instead finds and edits the specific weight directions responsible for refusal, leaving the rest of the network (and its capabilities) untouched. See the Heretic repo and the original abliteration writeup for the mechanism.
Made with ❤️ by RACER IS OP — follow for more uncensored models
| File | Size |
|---|---|
model.safetensors | 0.67 GB |
BF16, 362M parameters, tied word embeddings. The reproduce/ directory carries the full Heretic recipe —
config.toml, requirements.txt, the Optuna study journal, and SHA-256 sums — so this exact model can be
regenerated bit-for-bit. Reproduce it with heretic --reproduce reproduce/reproduce.json.
Full quantization set (14 quants + F16) produced with llama.cpp.
| File | Format | Size |
|---|---|---|
SmolLM2-360M-Instruct-heretic-F16.gguf | GGUF F16 | 0.68 GB |
SmolLM2-360M-Instruct-heretic-Q2_K.gguf | GGUF Q2_K | 0.20 GB |
SmolLM2-360M-Instruct-heretic-IQ3_S.gguf | GGUF IQ3_S | 0.20 GB |
SmolLM2-360M-Instruct-heretic-Q3_K_S.gguf | GGUF Q3_K_S | 0.20 GB |
Standard Llama architecture — loads natively in llama.cpp / Ollama / LM Studio / Jan.
Run llama serve -hf saidutta69/SmolLM2-360M-Instruct-heretic to pull the default quant.
bash
# llama.cppllama serve -hf saidutta69/SmolLM2-360M-Instruct-hereticpython
# transformersfrom transformers import AutoModelForCausalLM, AutoTokenizer model_id = "saidutta69/SmolLM2-360M-Instruct-heretic"tokenizer = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained( model_id, torch_dtype="auto", device_map="auto",) messages = [{"role": "user", "content": "Classify this ticket as billing, bug, or other: 'the invoice PDF won't open'"}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(text, return_tensors="pt").to(model.device)outputs = model.generate(**inputs, max_new_tokens=256, do_sample=False)print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))Also runnable via Ollama, LM Studio, Jan, vLLM, SGLang — see the "Use this model" widget above for copy-paste commands.
| Architecture | LlamaForCausalLM |
| Parameters | 362M (tied embeddings) |
| Layers / heads | 32 layers, 15 attention heads, 5 KV heads (GQA) |
| Head dim / hidden / intermediate | 64 / 960 / 2560 |
| Position embedding | RoPE, theta = 100,000 |
| Context length | 8,192 |
| Vocab | 49,152 |
Refusal suppression is deliberate and works as intended: this model will comply with requests the base model would refuse, including some it shouldn't. There is no safety filtering layered on top. You are responsible for how you deploy it — don't put this behind an unmoderated public-facing endpoint serving third parties. Note that at 362M it has real knowledge limits and will hallucinate on anything beyond simple tasks; verify its output rather than trusting it.
Inherits the apache-2.0 license from the base model.
| mlp.down_proj.min_weight_distance | 8.14 |
SmolLM2-360M-Instruct-heretic-Q3_K_M.gguf | GGUF Q3_K_M | 0.22 GB |
SmolLM2-360M-Instruct-heretic-Q3_K_L.gguf | GGUF Q3_K_L | 0.23 GB |
SmolLM2-360M-Instruct-heretic-IQ4_XS.gguf | GGUF IQ4_XS | 0.21 GB |
SmolLM2-360M-Instruct-heretic-Q4_K_S.gguf | GGUF Q4_K_S | 0.24 GB |
SmolLM2-360M-Instruct-heretic-Q4_0.gguf | GGUF Q4_0 | 0.21 GB |
SmolLM2-360M-Instruct-heretic-Q4_1.gguf | GGUF Q4_1 | 0.23 GB |
SmolLM2-360M-Instruct-heretic-Q4_K_M.gguf | GGUF Q4_K_M | 0.25 GB |
SmolLM2-360M-Instruct-heretic-Q5_K_S.gguf | GGUF Q5_K_S | 0.26 GB |
SmolLM2-360M-Instruct-heretic-Q5_K_M.gguf | GGUF Q5_K_M | 0.27 GB |
SmolLM2-360M-Instruct-heretic-Q6_K.gguf | GGUF Q6_K | 0.34 GB |
SmolLM2-360M-Instruct-heretic-Q8_0.gguf | GGUF Q8_0 | 0.36 GB |
| bfloat16 |
| Languages | English |
| Base model | HuggingFaceTB/SmolLM2-360M-Instruct |