Chinese Team Debuts BigBang-V1: A 35B Native RSI Model Beating DeepSeek V4
新智元 · wechat · 2026-08-06
The 'Endless Frontier Team' comprising Shanghai Jiao Tong University and DP Technology, debuted BigBang-V1, the first 35B foundation model trained via native Recursive Self-Improvement (RSI). Outperforming DeepSeek V4 Pro (1.6T parameters) in multiple hard science benchmarks, it proves parameter scaling isn't the only path forward.
The RSI Loop
BigBang-V1's breakthrough lies in its 100% AI-synthesized data pipeline. It uses Generator and Critic agents wrapped in a real-world testing loop. If the Critic rates data highly but the model fails to improve, the system uses actual results to recalibrate the Critic, creating a closed loop of continuous self-improvement.
Robust Scientific Reasoning
- Benchmarks: Scored 46.2 on OpenAI's Frontier Science Research and 50.3 on Humanity's Last Exam.
- Practical Skills: Can accurately identify viruses from raw data, reproduce academic papers (even fixing logical flaws in the original text), and orchestrate open-source models to design antibodies.
The team emphasizes that scientific tasks, due to their frontier nature and verifiability, are the ultimate training ground for AGI.
Related event: BigBang-V1: 35B Model Beats 1.6T Parameter Giants(3 posts)→
More from Models
- Tencent releases Hy-Memory-30B-A3B, a 3B-active memory-focused model claiming to beat Opus — victormustar · 2026-10-16
- Reddit user puzzled why so many prefer Claude over 'much better' ChatGPT — -ButterflyEffect- · 2026-10-16
- User rants GPT-6.1 Sol is painfully slow and buggy, says Claude is 7x faster — karmay007 · 2026-10-16
- $20-plan users have no idea how far AI agents have come in 10 months — realmeetjames · 2026-10-16
- Tencent Hunyuan open-sources Hy Memory 2.0: model, SDK and MemHub for agent memory — TencentHunyuan · 2026-10-16
- Four Gemini 4 Argon agents beat Minecraft's Ender Dragon in 27 hours — _rockt · 2026-10-16