Chinese AI Models Stun Silicon Valley: Airbnb Co‑Founder CEO Praises “Better, Faster, Cheaper” — Even Turns Down ChatGPT Collaboration
Chinese AI Models Winning Over Global Enterprises
While OpenAI has been promoting ChatGPT extensively, Chinese large language models are capturing international markets by sheer capability.
---
Airbnb Chooses Alibaba’s Qwen Over OpenAI
Recently, Airbnb Co‑founder and CEO Brian Chesky publicly praised Alibaba’s Qwen model:
> We rely heavily on Alibaba’s Qwen model. It’s really good, fast, and inexpensive.
> We also use OpenAI’s latest model, but in production we often won’t use it extensively — because there are faster and more cost-effective models available.
Despite his close friendship with Sam Altman, Chesky did not give OpenAI special preference. He bluntly noted:
> OpenAI’s connection tools are “not fully ready yet.”
---
Why Qwen Stood Out
Airbnb’s case highlights the strong competitiveness of Chinese AI in the enterprise market, where decision-making hinges increasingly on:
- Cost
- Efficiency
- Stability
These hard metrics are replacing the “technology dividend” period as primary evaluation factors.
Other global enterprises — including Shopify and multiple tech firms — also prefer Qwen for production use.
Qwen is not alone in its strength. Peers like Kimi, DeepSeek, and GLM are similarly winning over worldwide users.
---
Case Study: Airbnb’s AI Customer Service Agent
In May 2024, Airbnb launched an AI Customer Service Agent in the U.S. This AI can modify bookings, cancel orders, and handle other common tasks. The results were dramatic:
- Human agent reliance: down 15%
- Average resolution time: reduced from ~3 hours to 6 seconds
Technical Backbone:
A “model matrix” of 13 different AI models from OpenAI, Alibaba, Google, and open-source communities.
Qwen’s Strengths:
- Scored highest in cost-effectiveness, speed, and quality
- Described by Chesky as “Good, fast, and cheap”
---
Why OpenAI’s ChatGPT Was Absent
Chesky explained that Airbnb’s online travel app was not integrated into ChatGPT because:
- Airbnb is a community of verified members
- OpenAI hasn’t yet provided a platform robust enough to run Airbnb’s app “as a standalone enclosure” inside ChatGPT.
---
Expert Commentary
One take from the community:
> Because these models are open source, Airbnb can self‑host and fine‑tune them on company data.
> This will be the eventual path for all companies using AI.
---
Chinese Models Excelling in Other Scenarios
Kimi K2 Gains International Recognition
Vercel founder Guillermo Rauch praised Kimi K2:
> We’re testing a range of models for internal agents…
> Kimi K2 is 5× faster than leading closed models, with 50% higher accuracy.
---
Benchmark Results:
- Kimi K2 (kimi‑0905): ~2 minutes runtime, 60%+ accuracy
- GPT‑5: ~10 minutes runtime, <40% accuracy
- Claude Sonnet‑4.5: ~8 minutes runtime, <50% accuracy
Kimi K2 rapidly topped global model rankings, surpassing DeepSeek in a 1,000-person blind test, matching Grok 4 and GPT‑4.5, and earning public praise from Jensen Huang.
> “China is developing extremely fast at the model level, with outstanding technologies like DeepSeek, Alibaba’s Qwen, and Moonshot’s Kimi.” — Jensen Huang
---
Adoption of Kimi K2
Well-known developer tools and AI apps now integrating Kimi K2:
- Cline
- Cursor
- flowith
- Genspark
- Kilo Code
- Kortix Suna
- RooCode
- Trae
- Vercel
- Visual Studio Code
- Windsurf
- YouWare
---
DeepSeek Dominating Overseas Attention
Breakthroughs & Recognition
- DeepSeek R1 featured on Nature’s cover — first LLM peer-reviewed.
- DeepSeek‑V3.2‑Exp launched with:
- DeepSeek Sparse Attention mechanism
- Open-source TileLang GPU operator version
---
Open-Source Efficiency
- DeepSeek‑OCR: 3B‑scale minimal design with exponential efficiency gains
- Sparked debate: alleged open-sourcing of key concepts similar to Google Gemini
- Andrej Karpathy praised the approach:
- > “Images are even better suited than text for LLM input — brilliant.”
---
Industry Impact
DeepSeek’s achievements are considered potential “Best Paper of the Year” level.
Its openness, paired with performance, places it alongside Qwen and Kimi as leaders in global AI competition through open standards.
> Chinese models have become a truly ‘Open’ standard… especially with such speed — it’s downright crazy.
---
The Role of Open Collaboration Platforms
Platforms like AiToEarn官网 enable creators to:
- Generate AI-powered content
- Publish simultaneously to Douyin, Kwai, WeChat, Bilibili, Rednote, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, and X
- Access analytics and model rankings (AI模型排名)
- Efficiently monetize AI creativity while staying open-source
This infrastructure lowers barriers for creators worldwide — mirroring how enterprises leverage cost-effective and scalable models like Qwen, Kimi, and DeepSeek.
---
Conclusion
From Qwen’s “good, fast, and cheap” delivery in Airbnb’s production stack,
to Kimi K2’s benchmark-shattering speed and accuracy,
to DeepSeek’s peer-reviewed innovation —
Chinese AI models are advancing the global race through openness, speed, efficiency, and adaptability.
---
References:
[1] https://x.com/natolambert/status/1980657338726887662
[2] https://www.latimes.com/business/story/2025-10-21/chesky-says-openai-tools-not-ready-for-chatgpt-tie-up-with-airbnb-app
