DeepSeek Releases V4-Flash-0731 Open Model
The release continues DeepSeek's pattern of high-performance open models, with V4-Flash-0731 offering cost efficiency below the recently discounted GPT-5.6 Luna.
Reporting from 1 source: GIGAZINE.
DeepSeek has released DeepSeek-V4-Flash-0731 as an open model under a commercial-use license. The MoE model has 284 billion total parameters and 13 billion active parameters. Third-party tests by Artificial Analysis rate its intelligence on par with Google's Gemini 3.6 Flash, and it outperforms GLM-5.2 in coding and GPT-5.6 Luna in agent performance.
DeepSeek has published DeepSeek-V4-Flash-0731 on Hugging Face and ModelScope under the MIT License, making it freely available for commercial use. The model is an update to April's DeepSeek-V4-Flash and uses a mixture-of-experts architecture with 284 billion total parameters and 13 billion active parameters.
According to Artificial Analysis, the model matches Gemini 3.6 Flash in intelligence, trails only Kimi K3 and GLM-5.2 among open models, and beats GLM-5.2 in coding. Its agent performance exceeds Gemini 3.6 Flash by a wide margin and edges out GPT-5.6 Luna by 0.1 points. The model also stands out for its efficiency, costing less per task than the late-July discounted GPT-5.6 Luna while delivering comparable intelligence.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.