Japanese-focused lineage
The model card places it in the bakeneko series, which the card says is based on Qwen2.5 and continually pretrained on Japanese-specific corpora.
Open Source Model Profile · rinna
qwen2.5-bakeneko-32b-instruct
qwen2.5-bakeneko-32b-instruct is a 32.76B-parameter Qwen2-family instruction-tuned model from rinna. According to the model card, it builds on rinna/qwen2.5-bakeneko-32b with Chat Vector and SimPO.
- Publisher
- rinna
- Task
- text-generation
- Model type
- qwen2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
qwen2.5-bakeneko-32b-instruct is published by rinna as a Qwen2-based text-generation model. The captured configuration identifies Qwen2ForCausalLM and Safetensors metadata reports 32763876352 parameters. According to the model card, it is an instruction-tuned variant of rinna/qwen2.5-bakeneko-32b using Chat Vector addition and Simple Preference Optimization.
Recorded capabilities
Chat Vector plus SimPO
According to the model card, instruction ability was added through Chat Vector addition, then refined with SimPO.
Qwen2.5 chat format
According to the model card, the model adheres to the Qwen2.5 chat format and documents tokenizer.apply_chat_template use.
Publisher benchmark table
The model card reports Japanese LM Evaluation Harness and Japanese MT-Bench numbers, including 79.62, 8.17, and 7.66 for this model.
Use cases in the source record
- Japanese-language chat workflows using the documented Qwen2.5 chat format.
- Instruction-following experiments building on the publisher-described Chat Vector and SimPO lineage.
Limitations and unknowns
- No independently verified evaluation results were extracted; benchmark numbers above are publisher-reported and need validation.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: rinna/qwen2.5-bakeneko-32b-instruct
Captured: Unknown. Processed: 2026-09-07T19:35:29.595732+00:00.
Qwen2.5 Bakeneko 32B Instruct (rinna/qwen2.5-bakeneko-32b-instruct) Overview This model is an instruction-tuned variant of rinna/qwen2.5-bakeneko-32b , fine-tuned using Chat Vector and Simple Preference Optimization (SimPO). It adheres to the Qwen2.5 chat format and is designed to deliever superior performance in Japanese language tasks. Model Type Model Name Japanese Continual Pre-Training Model Qwen2.5 Bakeneko 32B [HF] Instruction-Tuning Model Qwen2.5 Bakeneko 32B Instruct [HF] [AWQ] [GGUF] [GPTQ int8] [GPTQ int4] DeepSeek R1 Distill Qwen2.5 Merged Reasoning Model DeepSeek R1 Distill Qwen2.5 Bakeneko 32B [HF] [AWQ] [GGUF] [GPTQ i…
F001F002F003F004F005F006F007F008F009F010F011F012F015F016F017F018F019F020F023