Tokenization Infinite
Category: AI Developer Tools
An enterprise-grade AI coding platform and multi-agent system designed to automate complex software development tasks, including repository-level issue resolution, code auditing, and task planning. Tokenization Infinite was founded in 2025. The company is led by Yang Ping. Based in Beijing, China. Team size: 30. Total funding raised: Tens of Millions RMB (Estimated). Latest round: Angel Round. Key investors include ["Confidential leading software industry CVC","Academic-linked investors from Tsinghua University network"].
- Founded
- 2025
- Headquarters
- Beijing, China
- Team size
- 30
- Total funding
- Tens of Millions RMB (Estimated)
Value proposition
Reduces software development cycles and costs by providing repository-level intelligence that can automatically fix bugs and generate patches with State-of-the-Art (SOTA) accuracy, while ensuring security through private cloud options.
Products and solutions
["InfCode Enterprise AI Coding Platform (Comprehensive R&D management)","InfCode IDE Extensions (Available on VS Code and JetBrains Marketplaces)","InfCode Open-source Code Agent System (Automated repository-level issue solver)","Private-deployment Security & Compliance Module"]
Unique value
Employs an 'Adversarial Iterative Refinement' (AIR) mechanism where multiple AI agents challenge and verify each other's code patches and test cases to ensure reliability.
Target customer
Enterprise R&D departments, large-scale software development teams, and organizations requiring secure, private-deployment AI coding solutions.
Industries served
["Software Development","Enterprise Information Technology","Financial Services (FinTech)","Cloud Computing & Infrastructure"]
Technology advantage
Achieved a world-leading 79.4% resolution rate on the SWE-bench Verified leaderboard, significantly outperforming competitors in repository-scale problem solving rather than simple line-level code completion.
How they differentiate
Employs a proprietary 'Adversarial Iterative Refinement' (AIR) mechanism where multiple AI agents iteratively challenge and verify code patches and test cases. Unlike snippet-level completion tools, it focuses on repository-level issue resolution, achieving a world-record 79.4% resolution rate on the SWE-bench Verified leaderboard (as of late 2025).
Main competitors
["Cognition AI (Devin)","ByteDance (Trae)","Zhipu AI (CodeGeeX)","GitHub (Copilot Workspace)"]
Key partnerships
["Confidential leading software industry Corporate Venture Capital (CVC) investors","Academic ties to Tsinghua University's Institute for Interdisciplinary Information Sciences (Yao Class)"]
Notable customers
["Confidential Enterprise R&D Departments","Domestic Financial Services Providers (POC stage)"]
Major milestones
["Company founded in July 2025 by former ByteDance Software Engineering Lab head Yang Ping.","Released 'InfCode' multi-agent system achieving 79.4% on SWE-bench Verified in November 2025.","Published foundational research 'InfCode: Adversarial Iterative Refinement of Tests and Patches' (arXiv:2511.16004).","Completed multi-million RMB Angel round within six months of founding."]
Growth metrics
Achieved SOTA (State-of-the-Art) performance on SWE-bench Verified (79.4%) using Claude 4.5 Sonnet as a backbone; transitioned from research to commercialization with the release of the InfCode Enterprise platform in December 2025.
Market positioning
High-end enterprise AI software engineering (AISE) platform challenger focusing on reliable, repository-scale automation.
Geographic focus
Headquartered in Beijing, China (Haidian District), focusing on the domestic enterprise market while maintaining a global technical presence through open-source benchmarks.
Patents and IP
Research published under the paper 'InfCode: Adversarial Iterative Refinement of Tests and Patches for Reliable Software Issue Resolution' (arXiv:2511.16004); proprietary multi-agent orchestration logic.
About Yang Ping
Former Head of the Software Engineering Laboratory at ByteDance and AI Technical Lead. Founded ByteDance's first AI-focused software engineering lab, where she developed internal AI coding products and multi-agent testing systems. Previously served as an engineer at Intel.
Official website: https://www.tokfinity.com