DeepSeek R1-0528 Now Supports Local Tool Calls on OpenRouter

AIbase基地

Published inAI News · 5 min read · Jun 10, 2025

The field of artificial intelligence continues to develop rapidly, and DeepSeek's latest update to its R1-0528 model has drawn significant attention from the developer community. According to AIbase, the upgraded DeepSeek R1-0528 now supports local tool calls on OpenRouter, a major improvement that enables developers to seamlessly integrate local tools into their AI workflows.

Local Tool Calls Enhance Workflow Efficiency

DeepSeek R1-0528 is an open-source inference model renowned for its performance comparable to proprietary models like OpenAI o3 and Google Gemini2.5Pro. This update introduces support for local tool calls via the OpenRouter platform, which serves as a unified gateway to access various AI models. This capability allows developers to connect the model with local tools such as code environments, data analysis tools, or custom APIs, simplifying the management of complex tasks and making R1-0528 an ideal choice for developers seeking cost-effective, high-performance AI solutions.

The introduction of local tool calls addresses a critical need within the AI ecosystem by combining advanced reasoning capabilities with practical, localized functionalities. With OpenRouter’s infrastructure, developers can now deploy R1-0528 across a variety of scenarios, from rapid prototyping to enterprise-level applications, without managing heavy computational resources.

Significance for the AI Community

The importance of this update lies in its democratization of AI capabilities. Unlike costly and access-restricted closed-source models, DeepSeek R1-0528’s open-source nature combined with OpenRouter’s free API service lowers the entry barrier for small organizations and individual developers. Recent discussions on social platforms highlight the enthusiasm among developers for this release, with many praising its excellent performance in deep reasoning and local tool integration.

In addition, this update further solidifies R1-0528’s outstanding performance. Benchmark tests show it excels in tasks like AIME2025 (accuracy increased from 70% to 87.5%) and LiveCodeBench (coding performance increased from 63.5% to 73.3%), rivaling industry leaders. The inclusion of the local tool call feature enhances its practicality, making R1-0528 a strong contender in the AI race.

Streamlined Model Improves Accessibility

Alongside the main model update for R1-0528, DeepSeek also launched a lightweight version, DeepSeek-R1-0528-Qwen3-8B, based on Alibaba’s Qwen3-8B model. This lightweight model significantly reduces computational resource requirements, running on a single GPU with 40GB to 80GB memory (such as Nvidia H100). Despite its smaller scale, it surpasses Google’s Gemini2.5Flash and OpenAI’s o3-mini in certain benchmark tests, making it an ideal choice for enthusiasts and developers with limited hardware resources.

This lightweight model is also available through OpenRouter, reflecting DeepSeek’s commitment to AI accessibility. Through OpenRouter’s free API layer, developers can experiment with R1-0528’s features without financial burden, fostering innovation on a global scale.

Apple Research Reveals the Truth Behind AI False Thinking: Reasoning Models Collapse Under Complex Tasks

A new study led by Apple researchers has poured cold water on the current large reasoning models (LRMs) that are highly anticipated. The research found that when solving complex tasks, reasoning models designed to simulate thinking processes, such as Claude3.7Thinking and Deepseek-R1, not only failed to show advantages but also exhibited serious problems like insufficient reasoning and performance collapse. This study tested four classic logical puzzles: the Tower of Hanoi, Checkers, River Crossing, and Block World. These puzzles allow precise control over the tasks.

Google releases preview version of Gemini2.5Pro: Coding performance beats DeepSeek R1 and Grok 3 Beta

Google recently released its latest preview version of Gemini2.5Pro, a product known as the 'most intelligent' model, which was first announced in March and then updated in May. The preview version aims to give enterprises an opportunity to test new applications or replace earlier versions. According to Google's official blog, this new version performs better in responsiveness and creativity, while surpassing other models in coding and reasoning ability. At the I/O Developer Conference in May, Google revealed that Gemini2.

Latest Internet Queen Report: Global AI Competition Landscape Changes, China and Open Source AI Breakthrough

Silicon Valley investor Mary Meeker's latest released AI report reveals that the global AI competition landscape is experiencing profound reshaping. Chinese strength and open source wave are shaking up the ecosystem dominated by top companies like OpenAI, entering a new stage of 'contested supremacy'. The report points out that China's large language models are rapidly catching up with the U.S., with models such as DeepSeek, Alibaba Qwen, and Baidu Wenxin Yiyan showing excellent performance in core tasks such as code generation and mathematical reasoning.

Epic Innovation! Fortnite Introduces Generative AI to Easily Customize Intelligent NPCs

At the recent 'State of Unreal' keynote, Epic Games made its first public demonstration of its latest technology that allows creators of Fortnite to use generative AI tools to customize personalized non-player characters (NPCs). The highlight of this feature is that developers can create free-roaming NPCs with unique personalities, voices, and tones using just about 20 lines of prompt text. According to reports from media outlet Engadget, the NPC in the demonstration was named

Huawei WATCH 5 Smartwatch Integrates with Two Large Models for Comprehensive Upgrade in Fitness and Health Experience

At today's Huawei Smart Wearable and Fitness New Product Technology Communication Conference, Huawei officially released its latest flagship product - the Huawei WATCH5 smartwatch. This watch not only continues to pursue fashion and technological feel in design, but also makes significant technical upgrades in functionality, especially in the fitness and health domain. The Huawei WATCH5 will integrate with two powerful artificial intelligence large models - the Pangu large model and the DeepSeek large model. The introduction of these models will enable the watch to enhance voice interaction and fitness and health capabilities.

Stanford's Latest Evaluation: DeepSeek R1 Medical AI Model Outperforms Google and OpenAI with High Scores

Recently, Stanford University released a comprehensive evaluation of clinical medical AI models. DeepSeek R1 stood out as the champion among nine leading large models, achieving a 66% win rate and a macro average score of 0.75. The highlight of this evaluation is that it not only focuses on traditional medical license exam questions but also delves into the daily work scenarios of clinical doctors, providing more practical assessments. The evaluation team developed an integrated assessment framework called MedHELM, which includes 35 benchmarks covering 22 subtasks in medicine.

DeepSeek may have used Google Gemini data to train new AI model

Recently, DeepSeek released an updated version of its latest R1 reasoning AI model, which demonstrated excellent performance in multiple math and coding benchmarks. However, DeepSeek did not disclose the source of the training data for its model, leading some AI researchers to question it and speculate that the model may have been partially trained based on Google's Gemini AI series. Melbourne developer Sam Paeach claimed that he found...