GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
Ismantic
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
Ismantic
Model Tree
Input Modalities
Output Modalities
Supported Functionality
S1 — S0 plus a parallel-data anneal. Branched from S0 at step 40,000 and ran the decay window (40,000 -> 45,149) on 1.2B tokens containing 30% Chinese-English parallel text.
S0 and S1 are a controlled pair: same starting checkpoint, same hyperparameters, same learning-rate schedule. Only the data differs.
This is a 13B-token model. For scale: Qwen3-0.6B-Base saw 36T tokens —
about 2,700x more. Treat the numbers below as what that budget buys, not as a
competitive result.
| WMT22 5-shot | BLEU | COMET |
|---|---|---|
| zh->en | 8.99 | 0.6855 |
| en->zh | 27.29 | 0.7743 |
Against S0 (0.54 / 3.97) this is more than an order of magnitude, and the jump is qualitative: S0 ignores the examples and gets the output language wrong, S1 actually translates. 1.2B tokens of anneal data — 30% of it parallel — is what made in-context learning appear.
The tokenizer is a compiled C++ extension, not loadable by
AutoTokenizer. The release ships tokenizer.py and example_load.py:
bash
pip install git+https://github.com/Ismantic/PieceTokenizerpython example_load.py
The model code (model.py, checkpoint.py) is bundled too — the package
depends only on torch plus the tokenizer extension, not on transformers.
Full pipeline, data mixes and every design decision (including the mistakes)
are documented in https://github.com/Ismantic/Summer. Notably
docs/WHY.md records why fp32 master weights are mandatory, why the learning
rate schedule is WSD rather than cosine, and what the vocabulary swap cost.
Apache-2.0. Training corpora are public datasets (FineWeb-Edu, Cosmopedia,
CCI3-HQ, SkyPile, WMT19, OPUS-100 and others; see data/source.py upstream).
Please observe their respective licenses.