AI热点 6小时前 119 阅读 0 评论

China's Baichuan Intelligence Launches Medical AI Model That Outperforms OpenAI in Key Benchmark

作者头像
AI中国

AI技术专栏作家 | 发布了 246 篇文章

Wang Xiaochuan, Founder and CEO of Baichuan Intelligence

TMTPOST -- After months of silence, Chinese AI startup Baichuan Intelligence, founded by former Sogou CEO Wang Xiaochuan, unveiled its latest large language model on Sunday, targeting the fast-growing AI healthcare market.

The new model, Baichuan-M2, is an open-source, medically enhanced LLM designed for private deployment in clinical settings. To address privacy concerns and cost constraints in healthcare, Baichuan says M2 is extremely lightweight — capable of running on a single RTX 4090 GPU after quantization — slashing deployment costs to about $1,400, or roughly 1/57th of DeepSeek-R1 H20’s dual-node setup. The model is also compatible with mainstream domestic chips, enabling hospitals to roll it out on existing hardware.

On the HealthBench medical benchmark, Baichuan-M2 scored 60.1, surpassing OpenAI’s gpt-oss120b (57.6) and other open-source leaders like Qwen3-235B and Kimi K2. In a harder subset of the benchmark, which OpenAI debuted with GPT-5 earlier this year, Baichuan-M2 scored 34.7 — making it the second model in the world to exceed the 32-point threshold.

“Building models for life and creating doctors for humanity is Baichuan’s mission,” Wang said, calling medical foundation models the “crown jewel” of AI research. He added that healthcare is the most representative industry for AI’s potential to “create people” — not just tools.

Founded in 2023, Baichuan has raised three funding rounds, including a $690 million Series A in July 2024, and is preparing a Series B at a $2.75 billion valuation. The company has already deployed AI pediatricians in Beijing Children’s Hospital and launched specialized models like the “Futang·Baichuan” pediatric foundation model.

The M2 model introduces an upgraded AI Patient Simulator and boosts processing speeds by 58.5% in critical care and outpatient scenarios compared with its predecessor. In clinical trials, M2 accurately diagnosed complex cases — such as hypothyroidism in a patient with persistent fatigue and neck swelling — based on nuanced symptom analysis.

Baichuan’s launch comes amid intensifying competition in AI healthcare, with ByteDance and Ant Group both rolling out standalone AI medical assistant apps in recent weeks. Frost & Sullivan projects China’s AI healthcare market will balloon from $1.2 billion in 2023 to $42.5 billion in 2033, a 43% annual growth rate.

作者头像

AI前线

专注人工智能前沿技术报道,深入解析AI发展趋势与应用场景

246篇文章 1.2M阅读 56.3k粉丝

评论 (128)

用户头像

AI爱好者

2小时前

这个更新太令人期待了!视频分析功能将极大扩展AI的应用场景,特别是在教育和内容创作领域。

用户头像

开发者小明

昨天

有没有人测试过新的API响应速度?我们正在开发一个实时视频分析应用,非常关注性能表现。

作者头像

AI前线 作者

12小时前

我们测试的平均响应时间在300ms左右,比上一代快了很多,适合实时应用场景。

用户头像

科技观察家

3天前

GPT-4的视频处理能力已经接近专业级水平,这可能会对内容审核、视频编辑等行业产生颠覆性影响。期待看到更多创新应用!