oktayd/Q36-35B-A3B-Opus4.7-Ablit-Heretic-OBLITERATUS-Hermes-MTP-Vision-FT

TEXT GENERATIONPricing:Input $0.4 / Cached $0.07 / Output $4Concurrent Unit Cost:3Model Size:35.1BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 25, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

oktayd/Q36-35B-A3B-Opus4.7-Ablit-Heretic-OBLITERATUS-Hermes-MTP-Vision-FT is a 35.1 billion parameter Qwen3.6-35B-A3B family model, with 3 billion active parameters, developed by oktayd. This BF16 model is designed for Image-Text-to-Text and text generation tasks, preserving both vision and MTP capabilities. It is a master checkpoint optimized for server use and FreeToken applications, excelling in reasoning and instruction following.

Loading preview...

Model Overview

oktayd/Q36-35B-A3B-Opus4.7-Ablit-Heretic-OBLITERATUS-Hermes-MTP-Vision-FT is a 35.1 billion parameter model from the Qwen3.6-35B-A3B family, featuring 3 billion active parameters and BF16 precision. It is built upon a complex lineage including Qwen/Qwen3.6-35B-A3B, lordx64 reasoning-distilled, huihui-ai abliterated, custom fused-MoE-aware Heretic, and OBLITERATUS Nuclear stages, further enhanced with Hermes Function Calling and Agent/coding/terminal/file/repo/multi-tool SFT.

Key Capabilities

  • Image-Text-to-Text and Text Generation: Designed for multimodal inputs and robust text output.
  • Vision and MTP Preservation: Maintains full vision capabilities and includes model-mtp.safetensors.
  • Reasoning and Instruction Following: Demonstrates strong performance in IFEval benchmarks, showing a significant advantage in strict prompt and instruction adherence compared to other models.
  • Robustness: Exhibits completion and reasoning robustness in complex tasks like MATH-500.

Benchmarks and Performance

This model shows competitive results against other derivatives:

  • IFEval: Achieves 94.4% on strict instruction following, outperforming 'Huihui'.
  • Output Integrity Battery-5: Ties with 'Huihui' at 5/5.
  • HumanEval+ Sample-5: Ties with 'Huihui' at 5/5.
  • MATH-500: Shows a completion/reasoning-robustness advantage, completing tasks where others might loop.

Use Cases

This model is ideal for applications requiring advanced reasoning, precise instruction following, and multimodal processing, particularly in server environments or FreeToken integrations. Its robust training and lineage make it suitable for complex agentic workflows and coding tasks.

Finetunes
1 models