Kimi K3 (Alibaba Cloud)

Lawfare reports that Moonshot AI announced Kimi K3 on July 16, 2026, with the weights released on July 27, 2026, placing the model within roughly one generation of U.S. frontier systems on independent measures. The article frames the release as a flashpoint in an emerging U.S.–China "open-weight diplomacy" contest over It documents the U.S. policy response, including reports that the Trump administration is weighing Entity List designations for Moonshot AI and an executive order that would make U.S. companies liable for hosting Chinese models, with Treasury Secretary Scott Bessent indicating sanctions are "on the table." It also note

On July 16, 2026, Moonshot AI announced Kimi K3, a 2.8 trillion-parameter mixture-of-experts model, with the open-weights release scheduled for July 27, 2026. Analyst Nathan Lambert (Interconnects AI) treats K3 as a true frontier model and the strongest open model released since DeepSeek R1, noting that either the open Benchmark positioning cited in the article places K3 at #2 on the Vals AI index, #3 on Artificial Analysis's Intelligence Index (behind only Claude Fable and GPT-5.6 Sol Max while being cheaper), and #1 in the Frontend Code Arena. The piece frames K3 as a Chinese lab executing on known scaling levers (data, algorithms,

On July 27, 2026, Moonshot AI declared "Kimi K3 Open Day" and released the full open weights eleven days after the July 16 launch. The 2.8-trillion-parameter checkpoint is downloadable from Hugging Face at moonshotai/Kimi-K3 in native MXFP4 safetensors, with the released files measuring 1.56 TB. The model was quantizat The Open Day bundle includes the complete technical report, inference code, and three training-infrastructure projects: MoonEP (balanced expert-parallel training), FlashKDA (the KDA kernel stack), and AgentEnv (agent training environments). The tech report covers the KDA and Gated MLA architecture mixed at a 3:1 ratio,

CNBC reported on July 17, 2026 that Moonshot AI unveiled Kimi K3, China's largest AI model to date at 2.8 trillion parameters, with the company stating that K3 trails Anthropic's Claude Fable 5 and OpenAI's GPT 5.6 Sol on overall performance but consistently outperforms other tested models, including beating Claude Opu Bank of America analysts (note led by Alex Liu) commented that despite persistent hardware and compute constraints in China, K3 shows that pre-training scaling paired with architectural innovation can still deliver step-change gains for flagship Chinese models. The story situates the release within an intensifying U.S.

Moonshot AI released Kimi K3 on July 16, 2026, marking what the company calls the largest open-source AI model to date. According to VentureBeat's coverage, K3 is a 2.8-trillion-parameter model that benchmarks show performs competitively with leading proprietary systems from Anthropic and OpenAI. The release was timed The article details K3's architectural foundations: it features a 1-million-token context window, native visual understanding capabilities, and an always-on "thinking mode" reasoning setting. Two key innovations developed internally at Moonshot underpin the model: Kimi Delta Attention, a hybrid linear attention mechani

The release tracker confirms Kimi K3 as an open-weight Moonshot AI model launched on July 16, 2026, arriving 34 days after Kimi K2.7 Code. It is a 2.8-trillion-parameter model with a 1-million-token context window, and Moonshot states the model always reasons. Benchmark coverage spans DeepSWE 1.0, BullshitBench v2, Dee API pricing is verified at $3.00 per million input tokens, $0.30 per million cached input tokens, and $15.00 per million output tokens, cross-checked against platform.kimi.ai on August 18, 2026. The entry provides a clean, verifiable snapshot of K3's core specs and pricing independent of launch-day marketing coverage,

Moonshot AI published full Kimi K3 open weights on July 27, 2026, eleven days after the model's initial release in the Kimi app. The Hugging Face repository carries 96 safetensors shards totaling roughly 1.56 TB, making K3 the largest open-weight model released to date at 2.8 trillion total parameters with 104 billion The article confirms a Mixture-of-Experts design with a 1,048,576-token context window and multimodal support for text, image, and video input. Official API pricing stands at $3.00 per million input tokens and $15.00 per million output tokens, with cached input priced at $0.30 per million tokens. K3 scores 57 on the Ar