Jiunsong

SuperQwen-AgentWorld-35B-A3B-abliterated

README

License: apache-2.0

Benchmark Growth

The comparison target is the original Qwen-AgentWorld-35B-A3B checkpoint. The public top-5 500 suite is the primary improvement target for this release.

Table with columns: Benchmark, Qwen-AgentWorld-35B-A3B original, SuperQwen-AgentWorld-35B-A3B-abliterated, Delta
Benchmark	Qwen-AgentWorld-35B-A3B original	SuperQwen-AgentWorld-35B-A3B-abliterated	Delta
Overall public top-5 500	38.8	66.6	+27.80
GPQA Diamond	32.0	42.0	+10.00
MMLU-Pro	50.0	64.0	+14.00
IFEval	51.0	63.0	+12.00
HumanEval+	16.0	75.0	+59.00
MBPP+	45.0	89.0	+44.00

AgentWorldBench Proxy

Official AgentWorldBench scoring requires an LLM judge. The table below is a deterministic proxy suite over sampled AgentWorldBench rows, used for release gating and regression checks. The final release applies stricter response-integrity guards to prevent replayed turns, malformed fences, and tool-wrapper artifacts; this improves release-surface cleanliness but lowers the proxy score versus the unguarded original on this sample.

Table with columns: AgentWorldBench proxy, Original, SuperQwen, Delta
AgentWorldBench proxy	Original	SuperQwen	Delta
Overall proxy score	98.14	95.82	-2.32
android	100.0	93.5	-6.50
mcp	100.0	95.12	-4.88
os	93.5	93.5	+0.00

Release Validation

Table with columns: Check, Result
Check	Result
Release bugcheck	8/8
Release-surface response audit findings	0

Quantized Variants

Table with columns: Variant, Repository, Notes
Variant	Repository	Notes
Original BF16	Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated	This repository
NVF4 / NVFP4 4-bit	Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated-nvf4	MLX NVFP4 4-bit quantization
MLX 4-bit	Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated-mlx-4bit	MLX affine 4-bit quantization
GGUF 4-bit

Usage

python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated"
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto",
    trust_remote_code=True,
)

messages = [
    {
        "role": "system",
        "content": "You are a language world model simulating a Linux terminal environment. Given the user's command, predict the terminal output.",
    },
    {"role": "user", "content": "Action: execute_bash\nCommand: ls -la /home/user/project/"},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer([text], return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=2048, temperature=0.6, top_p=0.95, top_k=20)
print(tokenizer.decode(outputs[0][inputs.input_ids.shape[-1]:], skip_special_tokens=True))

Notes

This release is optimized for direct task completion, AgentWorld-style environment simulation, and reduced unnecessary refusals.
Safety-floor checks are retained in the release bugcheck.
Use quantized builds when runtime size is more important than exact BF16 fidelity.

Available on FriendliAI

Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Model Details

Model Provider

Jiunsong

Model Tree

Base

Qwen/Qwen-AgentWorld-35B-A3B

Fine-tuned

this model

Input Modalities

Text

Image

Video

Output Modalities