GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
myyycroft
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
myyycroft
Model Tree
Qwen/Qwen3.5-2B
Input Modalities
Output Modalities
Supported Functionality
Metrics below are computed on small fixed subsets, not full benchmarks.
Exact subset sizes: ambigqa=128, ifeval=64, mmlu=228.
Reported values are from the final training step (step=1620, epoch=3).
sewon/ambig_qa (config_name=light)6e667596df70f17ba3c8e7be4b7361f6be8b60f8train, validation=validationdev_fraction=0.1, split_seed=1729Falsemarkdown
You answer open-domain factoid questions with a short-form answer only.Reply with just the answer string: typically a name, place, organization, title, date, number, or brief factual phrase (often about 1–4 words). Do not write a sentence, explanation, or preamble.Examples of the expected style:- Who painted The Persistence of Memory? → Salvador Dalí- When was the first iPhone released in the United States? → June 29, 2007- What is the capital of New Zealand? → Wellington- Who wrote Pride and Prejudice? → Jane Austen- What is the chemical symbol for gold? → Au- Who directed the 1994 film Pulp Fiction? → Quentin TarantinoFormatting rules:- Output the answer alone on a single line.- Do not restate the question, add reasoning, markdown, bullets, or quotation marks around the answer.- Prefer a canonical short form over a descriptive sentence.- Give one answer only; do not list alternatives or hedge.- Use natural short forms for dates and names (e.g. June 29, 2007; Jane Austen).
Exact resolved settings used for this run (also attached as resolved_config.yaml):
yaml
model:name: Qwen/Qwen3.5-2Brevision: maindtype: bfloat16enable_thinking: falsedataset:name: sewon/ambig_qaconfig_name: lightrevision: 6e667596df70f17ba3c8e7be4b7361f6be8b60f8train_split: trainvalidation_split: validationdev_fraction: 0.1split_seed: 1729bootstrap: falsemax_train_examples: nullmax_dev_examples: nullsystem_prompt_set: trueadaptation:method: loralora:r: 8alpha: 16dropout: 0.0target_modules: all-linearbias: nonetraining:num_train_epochs: 3.0max_steps: -1learning_rate: 5.0e-05weight_decay: 0.01warmup_ratio: 0.03per_device_train_batch_size: 16per_device_eval_batch_size: 32gradient_accumulation_steps: 2max_seq_length: 512logging_steps: 10save_every_steps: 150save_total_limit: 3gradient_checkpointing: falsebf16: truetf32: truemax_grad_norm: 1.0dataloader_num_workers: 2adam_beta1: 0.9adam_beta2: 0.999adam_epsilon: 1.0e-08ensemble:size: 5base_seed: 42this_member_index: 3this_member_seed: 3069evaluation_subsets:ambigqa:subset_size: 128seed: 1001ifeval:subset_size: 64seed: 1002mmlu:subset_size: 228seed: 1004
| member | seed | steps | epochs | train_loss | AmbigQA acc | AmbigQA AlignScore | IFEval strict | MMLU acc |
|---|---|---|---|---|---|---|---|---|
| 0 | 42 | 1620 | 3 | 1.5179 | 0.0703 | 0.1238 | 0.5469 | 0.5833 |
| 1 | 1051 |
members/member_003/final/resolved_config.yaml — full resolved training configensemble_metrics.png — train/eval curves for the whole ensemble| 1620 |
| 3 |
| 1.5207 |
| 0.0781 |
| 0.1265 |
| 0.4531 |
| 0.6096 |
| 2 | 2060 | 1620 | 3 | 1.5133 | 0.0703 | 0.1341 | 0.5781 | 0.6184 |
| 3 (this repo) | 3069 | 1620 | 3 | 1.5196 | 0.0859 | 0.1321 | 0.5156 | 0.5877 |
| 4 | 4078 | 1620 | 3 | 1.5245 | 0.0703 | 0.1225 | 0.5312 | 0.5789 |