Qwen2.5 Bakeneko 32B Instruct
Bakeneko Instruct combines the Japanese-adapted base with Qwen2.5's instruction chat vector and Simple Preference Optimization, or SimPO.
The instruction merge has an explicit formula
rinna adds the Qwen2.5 Instruct-minus-base difference with weight 1.0, excluding the embedding layer. Official AWQ, GGUF and GPTQ repositories distribute formats of the same release.Qwen2.5 Bakeneko 32B Instruct official model card
Keep the first instruction release distinct from V2
Practical guidance. Record the complete model identifier when comparing Japanese responses. A V2 file is not merely another quantization of this checkpoint, so mixing the names would make a comparison hard to reproduce.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | qwen2.5-bakeneko-32b-instructqwen2.5-bakeneko-32b-instruct-awqqwen2.5-bakeneko-32b-instruct-ggufqwen2.5-bakeneko-32b-instruct-gptq-int4qwen2.5-bakeneko-32b-instruct-gptq-int8 |
| License | Apache 2.0.Qwen2.5 Bakeneko 32B Instruct official model card |
| Prompt format | Qwen2.5 chat format.Qwen2.5 Bakeneko 32B Instruct official model card |
Related models
- Qwen2.5 Bakeneko 32B
rinna · Bakeneko
Bakeneko 32B continues pretraining Qwen2.5-32B on about 18B tokens from Japanese and English corpora.
Launch date not confirmedResearch releaseDownloadable - Qwen2.5 Bakeneko 32B Instruct V2
rinna · Bakeneko
A Japanese instruction model that merges Bakeneko with QwQ Bakeneko.
Launch date not confirmedDownloadable