joshycodes/qwen3-32b-control-SQa-sdf
TEXT GENERATIONPricing:Input $0.408 / Cached $0.0816 / Output $1.972Concurrent Unit Cost:2Model Size:32BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 23, 2026License:otherArchitecture:Transformer Featherless Exclusive Cold
joshycodes/qwen3-32b-control-SQa-sdf is a 32 billion parameter Qwen3 model, continued-pretrained by joshycodes on a self-authored corpus. This model was trained for one epoch on 32,354,045 tokens, focusing on a corpus the model itself generated based on its character and SDF principles. It represents a research checkpoint exploring self-authored data for continued pretraining, with a context length of 32768 tokens.
Loading preview...
Overview
joshycodes/qwen3-32b-control-SQa-sdf is a 32 billion parameter Qwen3 model that has undergone continued pretraining. This model is a research checkpoint, specifically designed to explore the impact of self-authored data on model development.
Key Characteristics
- Self-Authored Corpus: The model was continued-pretrained on a unique corpus that it generated itself. This corpus was created based on the model's understanding of its own character and the principles of SDF (Self-Describing Formats).
- Continued Pretraining: The training involved one epoch over 32,354,045 tokens, utilizing 40,448 documents, all of which were ordinary text, with none explicitly marked as self-authored in the provided README.
- Research Focus: This model is explicitly labeled as a research checkpoint. Its primary purpose is to investigate the effects of training on self-generated data.
Important Considerations
- Undeployed Status: The model has not been evaluated for capability, alignment, or identity. It is explicitly stated that this model should not be deployed for production use.
- Experimental Nature: This model is part of a broader research effort documented in the welfare-improvements repository, focusing on framing, planning, and evaluation of such experimental training methodologies.