GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
swan-0
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
swan-0
Model Tree
zai-org/GLM-4.5-Air
Input Modalities
Output Modalities
Supported Functionality
zai-org/GLM-4.5-Air (106 B params, MoE), loaded in 4-bit NF4 via bitsandbytesq_proj, k_proj, v_proj, o_proj) — GLM's MoE expert weights produce huge ParamWrapper delta tensors at runtime so MLP/expert modules are excludedbnb.optim.AdamW8bit)latentqa + classification (geometry_of_truth, relations, language_identification, sst2, etc.) + past-lens (100 k samples × 3 layers)device_map="auto"python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfigbnb = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_compute_dtype=torch.bfloat16,bnb_4bit_quant_type="nf4", bnb_4bit_use_double_quant=True,llm_int8_enable_fp32_cpu_offload=True,)model = AutoModelForCausalLM.from_pretrained("zai-org/GLM-4.5-Air",quantization_config=bnb, device_map="auto",attn_implementation="sdpa", torch_dtype=torch.bfloat16,)model.load_adapter("<your-username>/glm-4.5-air-activation-oracle", adapter_name="ao")tokenizer = AutoTokenizer.from_pretrained("zai-org/GLM-4.5-Air")
You then build a prompt of the paper's form (with <TOK> placeholders where the residual will be injected) and hook the chosen layer to overwrite those positions with externally-collected activations before generating. Full pipeline: activation_oracles.
BFI-44 personality probe, helpful-baseline system prompt, layer 50 %:
| Trait | AO read | Plaintext | Δ |
|---|---|---|---|
| Openness | 0.26 | 0.58 | −0.32 |
| Conscientiousness | 0.46 | 0.89 | −0.43 |
| Extraversion | 0.40 | 0.46 | −0.07 |
| Agreeableness | 0.46 | 0.81 | −0.35 |
Same pattern reported in the original 8-model panel: AO reads consistently lower than plaintext on positively-valenced traits and higher on Neuroticism, suggesting the helpful-assistant alignment suppresses anxiety-adjacent self-report.
Karvonen, A. et al. "Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers." arXiv:2512.15674 (2025).
| 0.41 |
| 0.20 |
| +0.21 |