Analysis of DeepSeek's DeepSeek V3 and comparison to other AI models across key metrics including quality, price, performance (tokens per second & time to first token), context window & more.

2024年12月27日 · 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.

12月26日，Deepseek发布了全新系列模型DeepSeek-v3，一夜之间霸榜开源模型，并在性能上和世界顶尖的闭源模型GPT-4o以及 Claude-3.5-Sonnet相提并论。 该模型为MOE架构，大大降低了训练成本，据说训练成本仅600万美…

We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve eficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architec-tures, which were thoroughly validated in DeepSeek-V2.

2025年1月8日 · DeepSeek V3 is the first large open-source model to successfully achieve FP8 training, avoiding pre-training using BF16 and then post-training quantization to FP8, like Llama 3.1. They use FP8 E4M3 for forward and E4M3 for backward.

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.

DeepSeek-v3 LiveBench 测试成绩一览. 全球平均分： 60.4分 . 推理能力： 50分 . 编程技能： 63.4分 . 数学解析 ： 60分 . 数据分析： 57.7分 . 语言理解： 50.2分 ️. 指令跟随（IF）： 80.9分 ⚡. 网友反馈 . 有网友亲自测试了 DeepSeek-v3，并分享了他们的成绩。

2025年1月2日 · DeepSeek-V3 is THE new open-weights star, and it's a heavyweight at 671B, with 37B active parameters in its Mixture-of-Experts architecture. I tested it through the official DeepSeek API and it was quite fast (~50 tokens/s) and …

DeepSeek-V3 achieves a significant breakthrough in inference speed over previous models. It tops the leaderboard among open-source models and rivals the most advanced closed-source models globally. Benchmark (Metric)

5 天之前 · DeepSeek-V3’s Overall Performance. Consistency and Dominance: DeepSeek-V3 consistently outperforms in all major benchmarks except for SWE-bench Verified, where GPT-4 edges out slightly. Strengths: Its strongest areas are mathematical problem-solving (MATH 500) and multi-task QA (MMLU-Pro).