We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks.
Notes: "At an economical cost of only 2.664M H800 GPU hours, we complete the pre-training of DeepSeek-V3 on 14.8T tokens, producing the currently strongest open-source base model. The subsequent training stages after pre-training require only 0.1M GPU hours." 6 * 37B (active params) * 14.8T = 3.2856e24 for pretraining. We know they trained in FP8. H800s get 1.513e15 FLOP/s in FP8: 2.688M * 3600 * 1.513e15 * MFU = 3.2856e24 Suggests a MFU of 0.2244 in pre-training. If we assume MFU was the same in post-training, that adds an additional: 0.1M * 3600 * 1.513e15 * 0.2244 = 1.222e23 FLOP from post-training Total: 3.2856e24 + 1.222e23 = 3.4078e24 FLOP
Size Notes: "We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities"
Notes: Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.