GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
NiuNiu0110
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
NiuNiu0110
Model Tree
Qwen/Qwen3.5-27B
Input Modalities
Output Modalities
Supported Functionality
| Base model | Qwen/Qwen3.5-27B |
| Training corpus | NiuNiu0110/RST-SFT-Qwen3.5-27B |
| Checkpoint | global_step_82 |
| Architecture | Qwen3_5ForConditionalGeneration |
| Weights | 51.7 GB, bfloat16, 1199 tensors |
| Vision tower | 333 tensors, copied verbatim from the base model |
Trained with verl + FSDP2 on a pre-tokenized corpus: the Qwen3.5 loss mask is baked into
the data once (scripts/15_export_pretokenized.py) rather than recomputed per backend,
because tokenizing turns separately and concatenating them does not reproduce the
whole-conversation render for this template.
None. No benchmark has been run against this checkpoint. The eval harness exists in the repo but has not been executed on these weights, so there is no score here to quote — not a low one, not a high one. Treat it as an untested artifact of a training run.
verl writes FSDP shards, which from_pretrained cannot load. These weights came from
scripts/08_prepare_eval_ckpt.sh, which merges every shard, splices the vision tower back
in from the base model (training carries only the text stack), and then verifies that the
text weights actually moved — a merge over a missing shard produces a loadable model that
is silently part-untrained, and a merge that reproduced the base is indistinguishable from
a successful one without that check.
python
from transformers import AutoModelForImageTextToText, AutoProcessormodel = AutoModelForImageTextToText.from_pretrained("NiuNiu0110/rst-qwen3.5-27b-sft", dtype="auto", device_map="auto")processor = AutoProcessor.from_pretrained("NiuNiu0110/rst-qwen3.5-27b-sft")
Apache-2.0, inherited from the base model.