DeepSeek-V3 distillation
According to the model card, the model distills DeepSeek-V3 with logit-level replication, documented as about 1.1B tokens of logits and elsewhere as an expanded 5B+ tokens of logits.
Open Source Model Profile · arcee-ai
Virtuoso-Medium-v2
Virtuoso-Medium-v2 is a Qwen 2.5 text-generation model from arcee-ai with about 32.76B parameters. According to the model card, it distills DeepSeek-V3 on top of a Qwen-2.5-32B base.
- Publisher
- arcee-ai
- Task
- text-generation
- Model type
- qwen2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Virtuoso-Medium-v2 is published by arcee-ai as a text-generation model. The captured configuration identifies Qwen2ForCausalLM with model type qwen2, and Safetensors metadata reports 32,759,790,592 parameters. According to the model card, it builds on Qwen-2.5-32B and distills DeepSeek-V3 with logit-level replication.
Recorded capabilities
Qwen 2.5 base with tokenizer work
According to the model card, the base is Qwen-2.5-32B, initially using the DeepSeek-V3 tokenizer for logit extraction and finishing with the Qwen tokenizer through specialized tokenizer surgery.
128k context claim
According to the model card, the context length is 128k tokens with a June 2024 knowledge-cutoff limitation.
Documented GGUF path
According to the model card, GGUF quantizations are available.
Use cases in the source record
- Advanced chat, enterprise analysis and automation, research simulation, language understanding, and STEM education uses listed by the publisher in the model card.
- Standard Transformers text-generation workflows following the card's documented tokenizer, generate, and decode example.
Limitations and unknowns
- No structured benchmark scores were extracted; BBH, MMLU-PRO, MATH, and comparison claims are publisher statements without extracted numbers.
- The distillation volume is inconsistently described in the card as both about 1.1B and 5B+ tokens of logits; both are unverified publisher claims.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
Source and provenance
Source: arcee-ai/Virtuoso-Medium-v2
Captured: Unknown. Processed: 2026-09-07T19:35:18.939405+00:00.
Virtuoso-Medium-v2 (32B) is our next-generation, 32-billion-parameter language model that builds upon the original Virtuoso-Medium architecture. This version is distilled from Deepseek-v3, leveraging an expanded dataset of 5B+ tokens worth of logits. It achieves higher benchmark scores than our previous release (including surpassing Arcee-Nova 2024 in certain tasks). GGUF Quantizations available here Model Details Architecture Base: Qwen-2.5-32B Parameter Count: 32B Tokenizer: Initially integrated with Deepseek-v3 tokenizer for logit extraction. Final alignment uses the Qwen tokenizer, using specialized “tokenizer surgery” for cross…
F001F002F003F004F005F006F007F010F011F012F013F014F015F016F017F018F020