GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
empero-ai
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
empero-ai
Model Tree
Qwen/Qwen3.5-4B
Input Modalities
Output Modalities
Supported Functionality
<think> block learned directly from Qwen3.8 2.4T A95B traces rather than synthetic self-generated reasoning.Measured with lm-evaluation-harness, HF backend, identical settings for base and student. Both models are reasoning models and are evaluated with the CoT protocols (gsm8k_cot, mmlu_flan_cot_zeroshot); MMLU covers all 57 subjects (~1,700 questions). Flexible-extract is the primary metric; strict-match requires exact answer formatting.
| Task | Metric | Qwen3.5-4B (base) | Qwen3.8-4B | Δ |
|---|---|---|---|---|
| gsm8k_cot | exact_match (flexible) | 0.850 | 0.785 | −0.065 |
| gsm8k_cot | exact_match (strict) | 0.850 | 0.785 | −0.065 |
| mmlu (CoT, 57 subjects) | acc (flexible-extract) | 0.354 | 0.553 |
Sampling for generation: temperature=0.6, top_p=0.95, top_k=20 (Qwen3.5 recommended settings).
python
from transformers import AutoModelForCausalLM, AutoTokenizerimport torchmodel_id = "empero-ai/Qwen3.8-4B"tok = AutoTokenizer.from_pretrained(model_id)model = AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.bfloat16, device_map="auto")messages = [{"role": "user", "content": "A snail is at the bottom of a 10-meter well. Each day it climbs 3 meters, each night it slips back 2. How many days until it escapes?"}]inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)out = model.generate(inputs, max_new_tokens=16384,temperature=0.6, top_p=0.95, top_k=20, do_sample=True)print(tok.decode(out[0][inputs.shape[1]:], skip_special_tokens=True))
A recent transformers release with Qwen3.5 support is required, along with the Gated DeltaNet kernels (flash-linear-attention and a CUDA-matched causal_conv1d build) — without them the linear-attention layers fall back to slow, memory-hungry PyTorch ops.
temperature=0.6, top_p=0.95, top_k=20. Greedy decoding on long generations is a known repetition-loop failure mode for reasoning models in this class.max_new_tokens (16,384 recommended); every answer opens with a <think> block. Parse and strip the <think>...</think> span for end users.Sign up for the Empero newsletter at empero.org for releases, evals, and research notes.
If this model helped you, consider supporting the project:
bc1qx6zepu6sfkvshgdmc4ewu6pk6rpadvpgffpp7vltc1qv2mefzps2vtjcpwfx8xxdrpplrcvltswm68r7xWeights are released under Apache-2.0, inherited from the Qwen3.5-4B base. Shared for research and experimentation, as-is.
| mmlu (CoT, 57 subjects) | acc (strict-match) | 0.071 | 0.233 | +0.162 |