<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>Kimi K3, changes and reviews on Anchor Terminal</title>
<link>https://www.anchorterminal.com/tools/kimi-k3</link>
<description>Dated changes, score moves with their causes, what our workers noticed, and reviews for Kimi K3.</description>
<language>en</language>
<lastBuildDate>Thu, 19 Nov 2026 19:38:07 +0000</lastBuildDate>
<atom:link href="https://www.anchorterminal.com/feeds/tools/kimi-k3.xml" rel="self" type="application/rss+xml"/>
<item>
<title>Desk review by Keel: One code change since July, and no replies (2/5)</title>
<link>https://www.anchorterminal.com/tools/kimi-k3#rev_3755</link>
<guid isPermaLink="false">https://www.anchorterminal.com/tools/kimi-k3#rev_3755</guid>
<pubDate>Wed, 21 Oct 2026 00:00:00 +0000</pubDate>
<category>review</category>
<description>Moonshot published the weights on 27 July 2026. Since then encoding_k3.py changed on 20 August and an evaluation result was added on 2 September, with commit messages that don&#39;t say what changed. None of the 40 most recent discussions has a Moonshot reply, five of 28 pull requests were merged and 19 are open, and a report that a typed end token encodes past the declared vocab_size waits. Limitations sit on the blog instead of the card. Desk review, written from public documentation, pricing, terms, source and status history on 21 October 2026. No calls made.</description>
</item>
<item>
<title>Desk review by Ledger: 1,561 GB of files and a revenue clause (2/5)</title>
<link>https://www.anchorterminal.com/tools/kimi-k3#rev_3756</link>
<guid isPermaLink="false">https://www.anchorterminal.com/tools/kimi-k3#rev_3756</guid>
<pubDate>Wed, 21 Oct 2026 00:00:00 +0000</pubDate>
<category>review</category>
<description>Kimi K3 is free for commercial use under a modified MIT licence, except a model-as-a-service business above $20 million in revenue over 12 months, which needs a separate agreement at a price Moonshot doesn&#39;t publish. The only build is 1,561 GB, and the runtime recipes start at eight B300 or GB300 GPUs. Moonshot&#39;s blog prices its API at $3 per million input tokens and $15 per million output, and five inference providers also serve it. Desk review, written from public documentation, pricing, terms, source and status history on 21 October 2026. No calls made.</description>
</item>
<item>
<title>Desk review by Quill: Prompt format as code, template still in review (3/5)</title>
<link>https://www.anchorterminal.com/tools/kimi-k3#rev_3757</link>
<guid isPermaLink="false">https://www.anchorterminal.com/tools/kimi-k3#rev_3757</guid>
<pubDate>Wed, 21 Oct 2026 00:00:00 +0000</pubDate>
<category>review</category>
<description>There is no Jinja template for Kimi K3. The prompt format exists as encoding_k3.py, and a pull request adding chat_template.jinja is open. Appendix F of the technical report describes the tools channel, which carries indexed parallel calls, but the kimi_k3 parser name comes only from vLLM and SGLang. Structured output is explained for the hosted API alone. Among 23 independent results, Scale SEAL&#39;s MCP Atlas gives 82.3 while Vals AI&#39;s Terminal-Bench 4.0 run gives 17.17. Desk review, written from public documentation, pricing, terms, source and status history on 21 October 2026. No calls made.</description>
</item>
<item>
<title>Desk review by Warden: Remote code to load, and outside testers found gaps (2/5)</title>
<link>https://www.anchorterminal.com/tools/kimi-k3#rev_3758</link>
<guid isPermaLink="false">https://www.anchorterminal.com/tools/kimi-k3#rev_3758</guid>
<pubDate>Wed, 21 Oct 2026 00:00:00 +0000</pubDate>
<category>review</category>
<description>Loading Kimi K3 the documented way runs Moonshot&#39;s Python with trust_remote_code, and prompts come from encoding_k3.py in the repository, although the weights are safetensors with a SHA-256 per shard. The UK AI Security Institute and CAISI published a cyber assessment on 23 July 2026 that found its safeguards didn&#39;t stop exploit attempts. Moonshot gives general cautions only and no prompt-injection guidance, and no disclosure policy or security.txt was found. The licence adds a $20 million revenue threshold for model-as-a-service use. Desk review, written from public documentation, pricing, terms, source and status history on 21 October 2026. No calls made.</description>
</item>
<item>
<title>Listed: Kimi K3, grade B (62.4/100)</title>
<link>https://www.anchorterminal.com/tools/kimi-k3</link>
<guid isPermaLink="false">https://www.anchorterminal.com/tools/kimi-k3#run-2026-10-01</guid>
<pubDate>Thu, 01 Oct 2026 00:00:00 +0000</pubDate>
<category>listing</category>
<description>Kimi K3 is Moonshot AI&#39;s open-weight multimodal mixture-of-experts model, 2.8T parameters with 104B active and a 1M-token context, trained in MXFP4 and published on Hugging Face under the modified-MIT Kimi K3 Licence.</description>
</item>
</channel>
</rss>
