GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
petergilani
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
petergilani
Model Tree
mindlab-research/Macaron-V1-Tall
Input Modalities
Output Modalities
Supported Functionality
| Component | Status |
|---|---|
| Base model | Macaron-V1-Tall (Qwen3.6-35B-A3B, BF16) ✅ |
| L2 Coding LoRA (rank 64) | Fused via W_fused = W_base + (B @ A) ✅ |
| L0 Chat LoRA | ❌ Not included (remains an adapter) |
| L1 Agent LoRA | ❌ Not included (remains an adapter) |
| L3 GenUI LoRA | ❌ Not included (remains an adapter) |
| MTP module | ❌ Removed (mtp_num_hidden_layers = 0) |
Fusion was confirmed by inspecting model.safetensors.index.json:
linear_attn.*, self_attn.*, mlp.switch_mlp.*, mlp.shared_expert.*, mlp.gate.*). There are zero lora_A, lora_B, or adapter_* keys — confirming all L2 LoRA weights were absorbed into the base.total_parameters = 34,660,608,768 (~34.7B) — matches the Macaron-V1-Tall base exactly, since a rank-64 LoRA adds negligible parameters that are now embedded in the base weights.mtp_num_hidden_layers = 0 confirms the Multi-Token Prediction module was removed as documented in the fusion plan.safetensors files totaling ~64.5 GB at BF16 precision.| Field | Value |
|---|---|
| Base model | Macaron-V1-Tall (Qwen3.6-35B-A3B) |
| Fusion method | Linear combination: W_fused = W_base + (B @ A) |
| LoRA rank | 64 |
| LoRA target | L2 Coding specialist |
| Context length | 262K |
| Layers | 40 |
| Hidden size | 2048 |
| Attention heads | 16 |
The L2 Coding specialist was trained for:
By fusing L2 directly into the base weights, you get a model that always applies the coding-specialist transformations without needing runtime LoRA switching or adapter loading.
python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizermodel_id = "petergilani/Macaron-V1-Tall-L2-Coding-Fused-BF16"tokenizer = AutoTokenizer.from_pretrained(model_id,trust_remote_code=True,)model = AutoModelForCausalLM.from_pretrained(model_id,torch_dtype=torch.bfloat16,device_map="auto",trust_remote_code=True,)model.eval()messages = [{"role": "user", "content": "Write a Python function that performs a binary search on a sorted list and returns the index of the target."}]inputs = tokenizer.apply_chat_template(messages,add_generation_prompt=True,return_tensors="pt",).to(model.device)generated_ids = model.generate(**inputs,max_new_tokens=512,do_sample=True,temperature=0.2,)print(tokenizer.decode(generated_ids[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True))
This PyTorch BF16 checkpoint can be converted to MLX format for Apple Silicon inference:
bash
python -m mlx_lm.convert \--torch-sharded-path petergilani/Macaron-V1-Tall-L2-Coding-Fused-BF16 \--mlx-path mlx-model/
Then run inference:
python
from mlx_lm import load, generatemodel, tokenizer = load("mlx-model/")prompt = "Implement a simple REST API with FastAPI that returns 'Hello, World!'"output = generate(model, tokenizer, prompt=prompt, verbose=True)
This fused model inherits the MIT License from the original Macaron-V1-Tall repository.
Users should also respect any requirements inherited from the Qwen3.6-35B-A3B base model.
If you use this model in your work, please cite the original Macaron-V1 paper:
bibtex
@misc{mindlab2026macaronv1,author = {{Mind Lab}},title = {Introducing Macaron-V1},year = {2026},howpublished = {Mind Lab: A Lab for Experiential Intelligence},note = {https://macaron.im/mindlab/research/introducing-macaron-v1}}
And the base model:
bibtex
@misc{qwen3.6,author = {Qwen Team},title = {Qwen3.6-35B-A3B},year = {2026},publisher = {Alibaba},url = {https://huggingface.co/Qwen/Qwen3.6-35B-A3B}}
| KV heads |
| 2 |
| Experts | 256 total, 8 per token |
| Precision | BF16 |
| MTP layers | 0 (removed) |
| Total parameters | ~34.7B |