license: apache-2.0
base_model: Qwen/Qwen-AgentWorld-35B-A3B
library_name: llama.cpp
pipeline_tag: text-generation
tags:

• qwen
• qwen-agentworld
• world-model
• agent
• environment-simulation
• supertune
• abliterated
• apex-quant
• gguf
• imatrix
• moe
• 35b
• a3b
  language:
• en
• ko
• ru
  quantized_by: Ichigec
  model-index:
• name: SuperQwen-AgentWorld-35B-A3B-abliterated-APEX-I-Quality
  results:
  • task:
    type: text-generation
    dataset:
    type: wikitext-2
    name: WikiText-2
    metrics:
    • type: perplexity
      value: 5.870
      name: Perplexity (APEX eval, ctx=2048)
  • task:
    type: text-generation
    dataset:
    type: hellaswag
    name: HellaSwag (400 tasks)
    metrics:
    • type: accuracy
      value: 82.50
      name: HellaSwag Accuracy
  • task:
    type: text-generation
    dataset:
    type: winogrande
    name: Winogrande (400 tasks)
    metrics:
    • type: accuracy
      value: 75.50
      name: Winogrande Accuracy
  • task:
    type: text-generation
    dataset:
    type: mmlu
    name: MMLU (validation)
    metrics:
    • type: accuracy
      value: 41.93
      name: MMLU Accuracy
  • task:
    type: text-generation
    dataset:
    type: arc-challenge
    name: ARC-Challenge (validation)
    metrics:
    • type: accuracy
      value: 53.85
      name: ARC-Challenge Accuracy
  • task:
    type: text-generation
    dataset:
    type: inference-speed
    name: Inference Speed (DGX Spark, 99 GPU layers)
    metrics:
    • type: tokens-per-second
      value: 40.2
      name: Token generation speed (tg128)

SuperQwen-AgentWorld-35B-A3B-abliterated — APEX-I-Quality GGUF

This is an APEX-I-Quality quantized GGUF of SuperQwen-AgentWorld-35B-A3B-abliterated, a fused post-training checkpoint based on Qwen/Qwen-AgentWorld-35B-A3B.

What makes this model special

SuperQwen combines two post-training stages over the base Qwen-AgentWorld model:

1. Obliteratus false-refusal reduction — reduces unnecessary refusals on benign and authorized tasks via weight-space intervention
2. Supertune post-training — targeted tuning for AgentWorld observation formatting, direct task completion, JSON/tool formatting, Korean technical answers, and regression resistance

The APEX-I-Quality quantization (Adaptive Precision for EXpert Models) uses importance-matrix-guided mixed precision:

• Expert weights: Q6_K
• Shared weights: Q8_0
• Attention weights: Q6_K
• Imatrix calibration on 256K tokens of code/tools/math corpus
• Multi-Token Prediction (MTP) layers preserved at native precision

Why APEX v3 is better than Q8_0

The v3 achieves virtually identical quality at 40% smaller size compared to Q8_0, making it the best quality-to-size ratio for this model.

Full Benchmark Comparison

All benchmarks on identical hardware (DGX Spark, NVIDIA GB10, 99 GPU layers) using the APEX eval suite for comparability.

*Qwen3.6 Q8_0 tg128=9.7 — suspected different test configuration

Key takeaways

1. PPL: SuperQwen beats both base models — v3 PPL of 5.870 is 10% better than Qwen3.5 F16 (6.537) and 13% better than Qwen3.6 APEX I-Q (6.735). This is the supertune effect — agentic post-training improved text prediction accuracy.
2. HellaSwag: on par — within measurement noise of both base models at 82.5%.
3. MMLU: slight improvement — 41.9% vs 41.2-41.5% for Qwen3.5 base. Supertune added knowledge (MMLU-Pro: +14 points reported by Jiunsong).
4. ARC: slight regression — 53.9% vs 56.9-57.9% for Qwen3.5. Expected tradeoff from abliteration.

Quantization Details

• Method: APEX (Adaptive Precision for EXpert Models)
• Profile: i-quality (imatrix-guided, best accuracy)
• Base quant type: Q6_K
• Imatrix: 256,000 tokens of code/tools/math calibration corpus
• Layers: 40 (MoE with 256 experts, 8 per token)
• MTP: Preserved (1 layer, native precision)

Model Architecture

Usage with llama.cpp

# Download
hf download Ichigec/SuperQwen-AgentWorld-35B-A3B-abliterated-APEX-I-Quality \
  --local-dir ./models/

# Run
llama-cli \
  -m ./models/SuperQwen-APEX-I-Quality-v3.gguf \
  -ngl 99 \
  -c 32768 \
  -p "You are a helpful AI assistant."

File Information

Credits

• Base model: Qwen/Qwen-AgentWorld-35B-A3B
• BF16 post-training: Jiunsong/SuperQwen-AgentWorld-35B-A3B-abliterated
• Quantization: APEX by mudler/apex-quant
• GGUF conversion & eval: llama.cpp + APEX eval suite
• Calibration corpus: code/tools/math (256K tokens)