Kimi K2

Evidence (18)

• 8Ars Technica AI·8/7/2026breakthrough
  ByteDance trains a giant AI model to challenge top US labs

  ByteDance is reportedly training an AI model with up to 10 trillion parameters, potentially putting it in the same class as Anthropic’s most advanced systems. The model is still in pre-training, and the final size may change before any fine-tuning or release.

• 7Hacker News·8/2/2026research
  Running Kimi K3 on AMD MI355X with Better Performance per Dollar Than B300

  The post reports benchmark results for serving the Kimi K3 model on AMD MI355X GPUs, claiming better performance per dollar than Nvidia B300 in this setup. It also describes a ROCm-side bug in speculative decoding infrastructure and notes that a small PyTorch fix was needed to stabilize the scheduler.

• 2Simon Willison·8/2/2026open_source
  Simon Willison publishes July 2026 newsletter preview

  Simon Willison posted an update announcing his sponsors-only monthly newsletter, with a preview of the June edition and a list of topics covered. The note highlights recent discussions of model behavior, several new model releases, and renewed interest in MCP, but it is primarily a newsletter announcement rather than a major product or research release.

• 4Simon Willison·7/31/2026research
  Simon Willison Discusses Open-Weight AI Models

  Simon Willison joined Bryan Cantrill and Adam Leventhal on the Oxide and Friends podcast to discuss a week of major AI news, centered on open-weight models competing with proprietary frontier systems. The conversation also covered AI security incidents, open-weight policy letters, and several broader digressions, with a note that some topics were already outdated by the time of publication.

• 5Hacker News·7/31/2026open_source
  Run Kimi K3 with 29 GB of RAM at 0.50 tokens per second

  A GitHub project shows a way to run the Kimi K3 model locally using about 29 GB of RAM, albeit at a slow 0.50 tokens per second. The Hacker News discussion focuses on the practicality of running a large model on modest hardware and the tradeoff between memory usage and inference speed.

• 7Hacker News·7/31/2026research
  Moonshot's Kimi reportedly runs on a 20,000-GPU Alibaba cluster

  A Hacker News discussion cites a report that Moonshot's Kimi is using a large Nvidia GPU cluster hosted by Alibaba. The post highlights the scale of the compute infrastructure behind the model, which suggests significant training or inference capacity.

• 5Hacker News·7/29/2026model_release
  Kimi K3 Gets a 256K Context Version

  Kimi Code has introduced k3-256k, a 256K-context variant of its Kimi K3 model. The update keeps the same results within the 256K window while using less quota than the 1M-context version, with guidance on how to switch models safely in coding tools.

• 6Hacker News·7/29/2026research
  Self-hosting Kimi K3 adds cost but improves task success

  A benchmark-style comparison tested self-hosted GPUs, rented hardware, and commercial APIs on 64 real coding tasks. The article finds that self-hosting can cost about 20% more than renting while achieving about 20% better task resolution in some setups, but matching top frontier-model performance requires very large GPU nodes.

• 4Hacker News·7/28/2026open_source
  Running Kimi K3 on an M1 Max

  The post shows an attempt to run Kimi K3 on an Apple M1 Max machine, highlighting practical local inference and performance considerations on consumer hardware. The discussion focuses on feasibility, efficiency, and community interest in running large models outside the cloud.

• 8Hacker News·7/28/2026model_release
  Kimi K3 Architecture Highlights LatentMoE and NoPE

  Kimi K3 is presented as a scaled-up production version of Kimi Linear, growing from 48B to 2.8T parameters and positioning itself as the largest open-weight model to date. The architecture emphasizes inference efficiency with components such as LatentMoE, Kimi Delta Attention, attention residuals, and a full switch to NoPE, while also adding native multimodal support.

• 3Hacker News·7/28/2026product_launch
  Open model use with a self-hosted endpoint

  The author describes setting up opencode to work with a personal inference endpoint on Modal and using Kimi K3, saying the experience felt simpler and more freeing than using hosted commercial AI plans. The piece is a first-person impression of the convenience and control of running an open model on infrastructure the user owns.

• 8Simon Willison·7/27/2026model_release
  Moonshot releases Kimi K3 open weights, but deployment is limited

  Moonshot AI has released the open weights for Kimi K3 on Hugging Face, making one of its largest language models available for self-hosting. The model targets long-horizon coding and knowledge work, but its 2.8-trillion-parameter MoE design and heavy hardware requirements mean only a small number of organizations can run it themselves.

• 7Hacker News·7/27/2026research
  Kimi-K3 Technical Report

  This item links to the technical report for Kimi-K3, with related discussion and an accompanying Hugging Face page. The report appears to focus on the model’s design, training, and evaluation details rather than a consumer product announcement.

• 7Hacker News·7/23/2026research
  Echo aims to match stronger AI systems with lower inference cost

  Echo is an experimental system that routes requests across a pool of open-weight models instead of using a single model for every task. The builder reports performance comparable to Fable on aggregate evaluations at about one-third the inference cost, while noting failures in allocation and combination decisions and ongoing testing on coding and agentic tasks.

• 7The New Stack·7/23/2026product_launch
  Cursor Launches a Model Router to Optimize Coding Requests

  Cursor has introduced a model router that routes each coding request to the model best suited for the task, balancing cost, speed, and capability. The company says early users saved 30-50% versus sending all requests to Opus 4.8, while Cursor also continues building its own in-house and frontier model stack.

• 5Hacker News·7/23/2026open_source
  Hacker News essay argues against anti-open-source AI claims

  The article argues that many objections to open source AI are overstated or flawed, and compares the debate to past fights over encryption and open-source software. It claims open source models are difficult to suppress and are supported by commercial incentives beyond China alone.

• 7Hacker News·7/21/2026research
  Kimi K3 Matches Fable on Agent Tasks, Routing Together Improves Results

  A benchmark study on about 1,030 real agentic tasks found Kimi K3 and Fable 5 performed at roughly similar quality overall, with each model showing strengths in different task categories. The post claims that routing between the two models achieved 93% accuracy and could cut costs by up to 50x versus using Fable alone on long agentic loops.

• 8Hacker News·7/20/2026model_release
  Kimi K3 and Qwen 3.8 Challenge Frontier Model Leaders

  The article says Moonshot Labs has launched Kimi K3 and Alibaba has launched Qwen 3.8, with both models reportedly nearing Anthropic’s Fable 5 in performance and their weights expected to be released publicly soon. It argues that open models closing the gap with top proprietary systems could intensify price and differentiation pressure on companies like Anthropic.