unconst/Affine-5czsc2fc98-r500-offline-dpo-hialpha-hirank-lobeta-softctx-extrasteps-merged

TEXT GENERATIONPricing:Input $0.4 / Cached $0.07 / Output $4Concurrent Unit Cost:3Model Size:35.1BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 16, 2026Architecture:Transformer Featherless Exclusive Cold

Affine-5czsc2fc98-r500-offline-dpo-hialpha-hirank-lobeta-softctx-extrasteps-merged is a 35.1 billion parameter model, derived from a LoRA merge of kevin954/Affine-5dfqbbh8ev-sft. This model is a salvaged checkpoint, indicating a focus on robust performance from an existing base. With a context length of 32768 tokens, it is designed for applications requiring extensive contextual understanding and processing. Its primary differentiator lies in its specific merging strategy, suggesting optimizations for particular tasks or data distributions.

Loading preview...

Model Overview

unconst/Affine-5czsc2fc98-r500-offline-dpo-hialpha-hirank-lobeta-softctx-extrasteps-merged is a 35.1 billion parameter language model with a substantial context window of 32768 tokens. This model is a LoRA-merged checkpoint, specifically salvaged from kevin954/Affine-5dfqbbh8ev-sft.

Key Characteristics

  • Parameter Count: 35.1 billion, offering a balance between capability and computational demands.
  • Context Length: 32768 tokens, enabling the processing of long documents and complex conversational histories.
  • Origin: It is a merged checkpoint, indicating a refinement or combination of prior training efforts, likely aimed at consolidating specific learned behaviors or improving overall performance from its base model.
  • Development Status: Described as a "salvaged checkpoint" and having "private TTL insurance," suggesting it's an internal or experimental iteration not yet cleared for broader public release or a final submission.

When to Consider This Model

Given its nature as a merged and salvaged checkpoint, this model is particularly suited for:

  • Research and Development: Exploring the effects of specific LoRA merging strategies on model performance.
  • Specialized Applications: If your use case aligns with the specific optimizations or data distributions that the base model kevin954/Affine-5dfqbbh8ev-sft was trained on, this merged version might offer enhanced capabilities.
  • Long Context Tasks: Its 32768-token context window makes it suitable for tasks requiring deep understanding of extensive textual inputs.