GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
ssyhw7
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
ssyhw7
Model Tree
Qwen/Qwen3.6-35B-A3B
Input Modalities
Output Modalities
Supported Functionality
| item | value |
|---|---|
| method | LoRA (Tinker managed), bf16 |
| rank / alpha | 32 / 32 (scaling 1.0; exported adapter_config lora_alpha=32) |
| target modules | all linear projections: self_attn q/k/v/o, Gated-DeltaNet in_proj_q/k/v/z + out_proj, shared_expert gate/up/down, routed experts w1/w2/w3 (fused 3D, one side shared across 256 experts), unembed (lm_head); router not adapted |
| data | Terminal / tool-use / research-execution trajectories (Terminal-Bench 1.x minus 2.0 overlap, AgentTuning, synthetic shell + science tasks) |
| examples / loss tokens | 2718 / 9,240,494 |
| steps / batch / LR / schedule | 170 / 32 / 0.0001 / cosine, 20 warmup |
| context length | 32768 |
| seed | 42 |
| checkpoint selection | lowest dev loss on a held-out split of the training mixture (0.2031 at step 170); benchmark data never used |
| Tinker source | tinker://423b1516-45ff-5480-b185-cc72ee3a4fa1:train:0/sampler_weights/e2_terminal_agent_r32_s42_20261016_step00170 |
Loss curves and raw metrics: curves/ in the code repository.
| source | location | licence | rows |
|---|---|---|---|
| terminal-bench-1-trajectories | mlfoundations-dev/terminal-bench-traces-local | unspecified on the HF card (tasks: Terminal-Bench, Apache-2.0; traces generated by the OpenThoughts/DCAgent team) | 42 kept of 80 fetched |
| terminus-swesmith-sandbox-traces | mlfoundations-dev/swesmith_with_plain_docker-sandboxes-traces-terminus-2 | unspecified on the HF card (underlying SWE-smith tasks MIT) | 387 kept of 3000 fetched |
| tmax-agent-terminus | NiuNiu0110/TMax-Agent-SFT-terminus | ODC-BY | 2999 kept of 3000 fetched |
| agenttuning | zai-org/AgentInstruct |
Contamination controls: SWE-bench Pro repos and instance ids, Terminal-Bench 2.0 task names,
ASI-Bench project ids excluded; HLE canary + 13-gram overlap filter. Details: data/README.md in the code repository.
See the code repository README; headline numbers are reported there with bootstrap confidence intervals. Per-benchmark results for this expert: pending (evaluation not yet run)
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.6-35B-A3B", revision="995ad96eacd98c81ed38be0c5b274b04031597b0", torch_dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(base, "ssyhw7/qwen36-35b-e2_terminal_agent")vLLM: vllm serve Qwen/Qwen3.6-35B-A3B --enable-lora --lora-modules e2_terminal_agent=<path to this adapter>
Trained on bandwidth-bounded samples of public datasets (see data README); no claim of improvement is made beyond what the evaluation tables show. Some upstream data (TMax) contains Gemini-generated content subject to Google's terms.
Code repository: https://github.com/HanyangWang0418-oss/qwen36-merge-takehome
| Apache-2.0 |
| 195 kept of 195 fetched |