DeepSeek open-sources DeepSeek-V2-Chat-0628 model code and improves mathematical reasoning capabilities

Recently, the Chatbot Arena organized by LMSYS released the latest list update. The LMSYS Chatbot Arena ranked 11th on the list, surpassing allOpen Source Model, including Llama3-70B, Qwen2-72B, Nemotron-4-340B, Gemma2-27B, etc., won the globalOpen SourceHonor for the top model.

DeepSeek- Compared with the 0507 open source Chat version, V2-0628 has comprehensively improved its capabilities in code mathematical reasoning, command following, role playing, JSON Output, etc.

DeepSeek open-sources DeepSeek-V2-Chat-0628 model code and improves mathematical reasoning capabilities

Chatbot Arena is a globally recognized authoritative large-model blind testing platform that uses manual blind testing to ensure the fairness of the evaluation. In this evaluation, DeepSeek-V2-0628 demonstrated world-class long-question solving capabilities in Hard Prompt, Code, Longer Query, and Math, and was at the same level as top models such as GPT-4-Turbo-0409 and Claude3Opus.

DeepSeek-V2-0628 not only performs well on the international stage, but also ranks among the best in the domestic model evaluation, ranking second among all domestic models, demonstrating its strong competitiveness. In addition, DeepSeek-V2-0628 was launched on June 28, 2024, providing API and web services at a very competitive price.

statement:The content is collected from various media platforms such as public websites. If the included content infringes on your rights, please contact us by email and we will deal with it as soon as possible.
Information

The U.S. Department of Commerce is spending $400 million to boost chip production

2024-7-20 8:22:56

Information

DeepL launches a new generation of translation AI, with translation performance surpassing GPT-4

2024-7-20 8:25:21

Cart
Coupons
Search