Nemotron-SFT Chat V2: Synthetic Dialogues from Qwen3 and GLM-4.6

NVIDIA released this synthetic dialogue dataset in March 2026 to improve model interactivity and instruction following. It contains multi-turn conversations generated by an ensemble of high-parameter models including Qwen3-235B, GLM-4.6, and Kimi-K2-Thinking.

Use Cases

• Supervised Fine-Tuning (SFT) using synthetic dialogues to improve conversational tone
• Instruction following training by leveraging precise prompts and responses from Qwen3-235B
• Reasoning capability enhancement using 'thinking' traces from Kimi-K2-Thinking outputs

Strengths

• Ensemble of 6+ high-parameter models including Qwen3-235B and GPT-OSS-120b
• Refreshed version of v1 with updated synthetic generation techniques
• Permissive ODC-By license allowing for commercial application

Limitations

• Synthetic data may contain hallucinations or logical inconsistencies inherent to the source models
• Lack of human-in-the-loop verification for the generated dialogues

Provenance

Source

NVIDIA

Collection Method

synthetic generation

Freshness

Updated March 2026