GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
lzwjava
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
lzwjava
Model Tree
Base
this model
Input Modalities
Text
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
Container
| Property | Value |
|---|---|
| Architecture | GPT-2 124M (12 layers, 12 heads, 768 hidden) |
| Parameters | 124,475,904 |
| Context Length | 1,024 tokens |
| Tokenizer | GPT-2 BPE (tiktoken) |
| Training Tokens | ~1.55B (1 epoch) |
| Training Steps | 47,000 |
| Validation Loss | 2.28 |
| Training Framework | nanoGPT |
| Training Hardware | NVIDIA RTX 4070 12GB |
| Training Time | ~8 hours |
| Bias | No (bias=False) |
SEC EDGAR filings sourced from the SEC-EDGAR corpus on HuggingFace, covering annual reports (10-K), quarterly reports (10-Q), current reports (8-K), and other filing types. Tokenized with GPT-2 BPE into ~1.55B tokens across 16 shards.
python
from transformers import GPT2LMHeadModel, GPT2Tokenizermodel = GPT2LMHeadModel.from_pretrained("lzwjava/sec-edgar-gpt-124m")tokenizer = GPT2Tokenizer.from_pretrained("lzwjava/sec-edgar-gpt-124m")prompt = "UNITED STATES SECURITIES AND EXCHANGE COMMISSION"input_ids = tokenizer.encode(prompt, return_tensors="pt")output = model.generate(input_ids, max_new_tokens=200, temperature=0.8, do_sample=True)print(tokenizer.decode(output[0]))
Training code and development notes: github.com/lzwjava/sec-edgar-gpt
bibtex
@misc{sec-edgar-gpt-124m,author = {Zhiwei Li},title = {SEC-EDGAR GPT-2 124M},year = {2026},publisher = {GitHub},url = {https://github.com/lzwjava/sec-edgar-gpt}}