Moonshot, a Chinese AI startup backed by Alibaba and Tencent, unveiled Kimi K3 on Friday, a 2.8 trillion-parameter model it claims is the world’s largest open-weight AI system, marking a major milestone in technology news as the company narrows the gap between open-source and proprietary frontier AI.
The model features a 1 million-token context window, native visual understanding, and a reasoning mode that allows it to process and retain substantially more information than earlier generations in a single prompt. According to Moonshot, Kimi K3 performed competitively with Anthropic’s Fable 5 and substantially outperformed OpenAI’s GPT 5.6 Sol and GPT 5.5 on several benchmarks measuring complex, multi-step tasks.
Third-party evaluations corroborate the company’s claims. On GDPval-AA v2, a benchmark measuring real-world tasks across 44 occupations and 9 major industries, Kimi K3 scored 1,687, placing it third overall behind only Fable 5 Max (1,815) and GPT-5.6 Sol Max (1,747.8), and ahead of Claude Opus 4.8 (1,600), according to Artificial Analysis. On Arena.ai’s Frontend Code Arena, the model claimed the top spot with a score of 1,679, outpacing both Fable 5 and GPT-5.6 Sol by a significant margin.
Kimi K3 is roughly 75 percent larger than DeepSeek’s V4 Pro, which held approximately 1.6 trillion parameters before K3’s release. The model incorporates two architectural innovations: Kimi Delta Attention, a hybrid linear attention mechanism, and Attention Residuals, a technique designed to improve scaling efficiency. Both were previously published as open research by Moonshot on GitHub.
Significance of Open-Weight Release and Developer Access
The decision to release full model weights on July 27 distinguishes Kimi K3 from proprietary systems locked behind API paywalls. Open-weight models allow users to download, run, and customize the underlying systems without being locked into vendor contracts. This approach has become a strategic priority for Chinese AI companies seeking to expand developer communities and global influence, according to Reuters reporting on the broader trend.
Moonshot is offering hosted API access at $3 per million input tokens and $15 per million output tokens—pricing positioned roughly in line with mid-tier Western offerings but at a performance level the company claims approaches the top of the market. Cached input tokens cost just $0.30 per million, and the company is running a promotional rebate through August 12 offering up to 30 percent back in vouchers for API credits of $1,000 or more.
The release comes as Chinese AI firms accelerate their model development cycles. Z.ai’s GLM-5.2, released earlier in 2026, stunned industry observers by scoring near top U.S. closed-source models on benchmarks, undermining a Western consensus that Chinese AI models lagged by at least six months. MiniMax is also developing its own 2.7-trillion parameter model planned for release by the third quarter of 2026.
Moonshot’s Recovery and Market Position
Kimi K3 represents a remarkable turnaround for Moonshot. Founded in 2023 by Yang Zhilin, a Tsinghua University graduate who previously conducted research at Google and Meta, the company quickly became one of China’s most prominent AI startups. By early 2026, it had raised roughly $1.5 billion across multiple rounds, with its valuation climbing to $4.3 billion.
Then DeepSeek’s low-cost R1 model disrupted the Chinese AI landscape in January 2025. Moonshot’s Kimi platform, which had ranked third in monthly active users in China, slid to seventh. The company’s strategic pivot to open-source models—beginning with Kimi K2 in July 2025 and accelerating with K2.5 in January 2026—was an effort to reclaim relevance. Training a 2.8-trillion-parameter model requires enormous computational resources and months of preparation, suggesting Moonshot locked in architectural decisions well before the model reached the public.
The timing of the release, coinciding with the 2026 World Artificial Intelligence Conference in Shanghai, underscores the geopolitical dimensions of the open-source AI movement. By releasing the world’s largest open-source model, Moonshot is positioning itself as a center of gravity for the global open-source AI developer community—a strategy that helps China counter U.S. efforts to limit Beijing’s technological progress through export controls on advanced semiconductors.
Sources
- Reuters — Reported Moonshot’s Kimi K3 unveiling, benchmark performance, architectural innovations, and the company’s backing by Alibaba and Tencent
- VentureBeat — Provided detailed technical specifications, benchmark data from Artificial Analysis and Arena.ai, pricing information, and context on Moonshot’s market recovery
- The Hindu — Confirmed July 27 release date for full model weights and Moonshot’s backing by Alibaba and Tencent
- South China Morning Post — Reported on the model’s capabilities and China’s narrowing gap with U.S. rivals
- The Decoder — Provided benchmark analysis comparing K3 to Fable 5 and GPT-5.6 Sol
