CoCreate 2026: Scale Surge, US SMEs Truly Arrive
On September 9, 2026, AliExpress Trade’s annual overseas buyer conference CoCreate 2026 opened at the Los Angeles Convention Center. Key facts:
- Date: September 9, 2026 (same day as Apple’s product launch)
- Location: Los Angeles Convention Center, USA
- Attendees: Over 15,000 US small and medium-sized enterprises (15,000 SMEs), up from 3,000 in 2025
- New announcement: Accio, AliExpress Trade’s AI workbench, open-sourced the first e-commerce Agent benchmark test on GitHub
The growth is staggering—a fourfold year-on-year increase, revealing a sharp uptick in local enthusiasm for sourcing via Chinese supply chains.
From B2B to A2A: American Shop Owners Learning to Partner with AI
CoCreate this year was described as a gathering where small business owners sought “the next big opportunity in the AI era.” Moving beyond traditional B2B (business-to-business), the event emphasized the rise of A2A (Affiliate-to-Affiliate): independent store operators, content creators, and micro-distributors directly engaging Chinese suppliers—empowered by AI tools for rapid product selection, fulfillment, and customer service.
This shift reflects a deeper structural change in US commerce. As platforms like Shopify and Walmart Marketplace intensify competition, solo operators need lighter, smarter tools. The surprise stats: while AliExpress Trade historically attracted large export firms, 15,000 attendees included mostly micro-enterprises with fewer than 50 employees. This signals a newfound penetration into the U.S. “mom-and-pop shop” ecosystem.
Booth(s) become real-time market sensors for Chinese sellers. Many reported US buyers now focus less on pricing and more on “Does it integrate with local AI customer service?” or “Can it sync with TikTok Shop workflows?”
Accio’s开源 (Open-Sourcing): 50% Token Savings Across 44 Tasks
AliExpress Trade simultaneously announced that its global e-commerce AI platform Accio open-sourced the first real-world e-commerce Agent benchmark test on GitHub.
The benchmark spans 44 core tasks—product listing, order fulfillment, multilingual support, promotion management—reflecting actual transaction flows. Compared to generic large-model Agents, Accio’s approach slashes token consumption by 50%, highlighting its specialization for vertical commerce use cases.
| Metric | Accio E-commerce Agent | Generic Agent (Reference) |
|---|---|---|
| Scope | 44 real e-commerce tasks (full transaction simulation) | Mixed-domain general tasks |
| Token cost | 50% lower | Baseline |
| Availability | Open-source on GitHub | Often proprietary |
| Target users | Independent stores, cross-border sellers, SaaS providers | Broad AI developer community |
Note: Data sourced solely from original material; generic Agent benchmark figures are industry-typical.
Notably, lower token usage does not necessarily imply smaller model size. Accio’s efficiency may stem from tree-structured reasoning pruning, task-chaining optimization, and domain-specific knowledge injection—a compute-first trade-off rather than capability-first.
Who Should Act, and Who Should Wait?
- Ready to adopt now: Independent-store sellers with annual GMV of $500K–$5M and an in-house tech team. Accio’s open benchmark serves as a reflexive baseline to compare vendor proposals.
- Best to wait: Ultra-small operations (1–3 people) lacking dedicated service staff. Native-agent robustness and post-sale fallback mechanisms remain unclear pending public testing.
Watch the task taxonomy carefully—how closely do the 44 tasks mirror your actual workflow? That alignment dictates benchmark relevance.
Final Thought
15,000 attendees signal a pivotal shift: as hype around general-purpose LLMs cools, agents that actually run profitable commerce flows—cheaply, reliably, at scale—are winning both developer interest and real-world adoption.)