發生什麼事
Together AI says on @togethercompute that post-training is now available for DeepSeek V4.1 Flash (version 1017). You can run supervised fine-tuning (SFT) or reinforcement learning (RL) with your own data, rewards, and environments, while keeping the training loop in your own Python code. The company positions this as a way for developers to adapt the model for coding and agent tasks.
01Sources
Evidence: 🗣 Official self-report; not independently tested.
- @togethercompute: Together AI's official X post, published 2026-11-18 04:00 UTC (12:00 Taipei time).