unconst/Affine-5czsc2fc98-r487-offline-dpo-hialpha-lolr-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

The unconst/Affine-5czsc2fc98-r487-offline-dpo-hialpha-lolr-extrasteps-merged model is a 35.1 billion parameter language model, derived from a LoRA merge of kevin954/Affine-5dfqbbh8ev-sft. It features a substantial context length of 32768 tokens, making it suitable for processing extensive inputs. This model is specifically noted as a private TTL insurance checkpoint, indicating its origin from a specific development stage rather than a general release. Its primary application is within a specialized, internal development context, likely for further refinement or evaluation.

Loading preview...

Model Overview

The unconst/Affine-5czsc2fc98-r487-offline-dpo-hialpha-lolr-extrasteps-merged is a large language model with 35.1 billion parameters and a context window of 32768 tokens. This model is a result of a LoRA merge, specifically originating from kevin954/Affine-5dfqbbh8ev-sft.

Key Characteristics

  • Parameter Count: 35.1 billion, indicating a robust capacity for complex language understanding and generation tasks.
  • Context Length: A significant 32768 tokens, allowing for the processing and retention of information from very long inputs.
  • Origin: It is a merged checkpoint, salvaged from a specific development phase, and is described as "private TTL insurance." This suggests it's an internal or interim version, not intended as a public submission until further development stages are cleared.

Intended Use

This model is currently designated as a private checkpoint, likely for internal evaluation, testing, or as a backup during a multi-stage development process. Its status as "private TTL insurance" implies it serves as a safeguard or intermediate state within a larger project, rather than a model ready for broad deployment or general use cases. Developers might use this version for specific internal benchmarks or to continue development from a stable, salvaged point.