DeepSeek-R1-0528

Claims and evidence

• Released 28 May 2025 ● Primary source[3]news sidebar: DeepSeek-R1-0528 Release 2025/05/28 [2]Date: 2025-05-28 [4]page date

• Status AvailableWeights are still public on Hugging Face. In the DeepSeek API, deepseek-reasoner moved to the thinking mode of DeepSeek-V3.1 on 2025-08-21. ● Primary source[1] [2]Date: 2025-08-21
• Replaced by DeepSeek-V3.1In the DeepSeek API only: deepseek-reasoner was upgraded to DeepSeek-V3.1 (thinking mode). The weights stay available. ● Primary source[2]Date: 2025-08-21

• Revision of DeepSeek-R1The model card calls it a minor version upgrade of DeepSeek R1. ● Primary source[1]1. Introduction [3] [2]Date: 2025-05-28

• Change · Tool use JSON output and function calling support are listed as part of the upgrade.Compared with DeepSeek-R1 ● Primary source[2]Date: 2025-05-28 [3]
• Change · Reasoning Deeper reasoning, which DeepSeek attributes to more compute and algorithmic optimisation in post-training; complex tasks may use more tokens.Compared with DeepSeek-R1 ● Primary source[1] [2]Date: 2025-05-28
• Change · Other Fewer hallucinations, according to DeepSeek.Compared with DeepSeek-R1 ● Primary source[3]
• Change · Other Supports a system prompt and no longer needs to be forced into its thinking pattern at the start of the output.Compared with DeepSeek-R1 ● Primary source[1]

• Feature Function callingThe release notes list JSON output and function calling support as part of this upgrade. ● Primary source[3] [2]Date: 2025-05-28
• Open weights YesMIT License; commercial use and distillation supported. ● Primary source[1]5. License [3]
• Access consumer app, API, Open-weights download ● Primary source[3] [1]3. Chat Website & API Platform

Lineage

Variants

No variants recorded in this record.

Related AI Radar coverage

AI Radar coverage starts in June 2026; no coverage linked yet.

All model releases from DeepSeek on AI Radar →