GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
pinkelephantlimited
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
pinkelephantlimited
Model Tree
Input Modalities
Output Modalities
Supported Functionality
| Parameter | Value |
|---|---|
| Architecture | LLaMA (decoder-only) |
| Parameters | 22,026,624 |
| Hidden size | 384 |
| Intermediate size | 1,536 |
| Num layers | 8 |
| Num attention heads | 12 |
| Num KV heads | 12 |
| Max sequence length | 1,024 |
| Vocabulary size | 8,192 |
| Position encoding | RoPE (θ=10000.0) |
| Activation | SwiGLU |
| Normalization | Pre-RMSNorm |
| Weight tying | Yes (embedding ↔ lm_head) |
| Precision | fp32 |
| Source | Type | Size |
|---|---|---|
| HumanEval | Code problems | ~164 examples |
| LAMBADA | Narrative text | ~5,000 examples |
| Parameter | Value |
|---|---|
| Hardware | 15GB VRAM |
| Optimizer | AdamW |
| Learning rate | 1e-4 |
| Batch size | 128 |
| Epochs | 10 |
| Training time | ~2 minutes |
This model was trained entirely from scratch — no fine-tuning, no transfer learning, no derivative weights. Every parameter is original.
MIT
At 22M parameters, this model runs on Raspberry Pi, mobile browsers via WebAssembly, or any edge device:
python
from transformers import pipelinepipe = pipeline("text-generation", model="pinkelephantlimited/pink-elephant-22m")# Real-time code suggestion in a lightweight editorpipe("def fibonacci(n):", max_new_tokens=40)pipe("import pandas as pddf = pd.read_csv('data.csv')df.", max_new_tokens=40)
python
# Complete partial sentences on IoT devices or smart keyboardspipe("The quick brown fox jumps", max_new_tokens=20)
python
# Auto-generate code suggestions in GitHub Actions# Runs in <100ms on CPUpipe("class UserModel:", max_new_tokens=60)
Why 22M? Loads in milliseconds, runs on any hardware, uses <100MB RAM. Ideal for real-time applications where latency matters more than fluency.