Gate News message, April 27 — Moonshot AI plans to launch Kimi K3 in Q3 2026 (July-September), according to reports. The model will feature over 2.5 trillion parameters, with internal testing revealing context window capabilities exceeding 1 million tokens.
However, whether the 1M context will be available in the public release remains uncertain. The primary bottleneck limiting the 1M context rollout is computational resources rather than technical capability, sources indicate. Kimi’s current public version K2.6 maintains a 256K context window.
The development comes as DeepSeek V4 has already positioned 1M context as a public offering, raising the bar for long-context capabilities in open-weight models. If Kimi K3 brings 1M context to its public release, Moonshot would match DeepSeek V4 in both model scale and long-context performance.
Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to
Disclaimer.
Related Articles
Google DeepMind joins South Korea’s K-Moonshot to tackle scientific challenges with AI
Google DeepMind has entered into a partnership with South Korea to provide the technical backbone for the country's K-Moonshot initiative, an ambitious national programme designed to solve the most complex scientific challenges of the century using AI.
Summary
Google DeepMind partners with Sout
Cryptonews8m ago
Gate to Host AI Trading Space Roundtable on April 28: Exploring AI as the Next Web3 Cycle Driver
Gate News message, April 27 — Gate will host a live Space roundtable discussion on AI Trading on April 28 at 8 p.m., bringing together industry experts to explore whether AI's deep integration into trading workflows marks the true starting point of the next Web3 cycle.
The discussion will examine A
GateNews12m ago
Mac Studio Runs Large-Model Performance Tests: M3 Ultra, Cluster Solutions, and Expectations for M5 Ultra
In April 2026, models at the trillion-parameter class such as DeepSeek V4 Pro and Kimi K2.6 were released in succession, making “running frontier open-source LLMs on your own machine” a feasible option. For engineers and small teams who don’t want to build an H100 workstation but still want full local inference capability, \\Mac Studio M3 Ultra 256GB\\ is currently the most cost-effective single-machine solution, and with a Thunderbolt 5 cluster it can go up into the 1T-parameter range. This article compiles real-world test data for running large models on the M3 Ultra, cluster solutions, the advantages of the MLX framework, and the expected timeline for the M5 Ultra.
Current M3 Ultra specs: 256GB unified memory, 819
ChainNewsAbmedia14m ago
Musk Sues OpenAI, Microsoft Over Nonprofit Mission Shift
Jury selection is set to begin in Elon Musk's 2024 lawsuit against OpenAI, Sam Altman, Greg Brockman, and Microsoft, accusing them of betraying OpenAI's nonprofit mission by creating a for-profit entity in 2019, according to Reuters. Musk is seeking US$150 billion in damages for OpenAI's
CryptoFrontier16m ago
Intel's Stock Surges 110% as AI Transition Reshapes Tech Industry
Gate News message, April 27 — Intel's stock has climbed 110% this year and reached a new all-time high on Friday, marking a significant turnaround for the chipmaker 25 years after its previous peak. The resurgence reflects a broader shift in the technology sector, where the AI transition is
GateNews26m ago
Ant Group Launches Ling-2.6-1T: Trillion-Parameter Model Optimized for Token-Efficient Task Execution
Gate News message, April 27 — Ant Group's inclusionAI has released Ling-2.6-1T, a new trillion-parameter flagship instruction model in the Ling series. Unlike long-chain reasoning models, Ling-2.6-1T employs a "Fast-Thinking" mechanism designed for precise task execution with minimal token
GateNews39m ago
