Kimi / Moonshot AI News

Latest reporting, research and product updates filed under Kimi / Moonshot AI.

34 picksNewest firstLatest article Sep 24, 2026, 8:00 AM

Latest stories

September 28, 20264 stories

Fireworks AI Releases Ember-1: A Post-Trained Kimi K3 That Uses About 40% Fewer Tokens

Fireworks AI has released Ember-1, a specialized model from Fireworks Research built by post-training Moonshot AI’s open-weight Kimi K3. Ember-1 learns to produce shorter reasoning traces while keeping task accuracy. This is different from lowering the reasoning effort setting at inference time. According to the Fireworks release post, Ember-1 delivers Kimi K3’s quality with about 40% fewer tokens. Is it deployable? Yes, but only through the Fireworks serverless API as a Research Preview. Firewo

06:43X：Elvis Saravia (@omarsar0, DAIR.AI)AI score 55/100

This is a bigger deal than it seems. I like this push on the Pareto frontier to squeeze as much as ...

This is a bigger deal than it seems.I like this push on the Pareto frontier to squeeze as much as you can out of your tokens.Feels underexplored.All labs are quick to launch models, so certain aspects just aren't optimized.These are just a few of the great things you can start doing with frontier open models.Ember-1 produces shorter reasoning traces (40% fewer tokens) without sacrificing performance.

Ember-1

Ember-1: half the tokens, same answersEmber-1 is a new specialized model from Fireworks Research that delivers Kimi K3’s quality with 40% fewer tokens. Built on Kimi K3, it learned to cut unnecessary reasoning while keeping the thinking that matters. We tested it on external benchmarks, in live customer A/B tests, and on our own coding and agent workloads, and quality held up in every setting. Available today, Ember-1 kicks off an ongoing series of specialized models by Fireworks, shaped by what

01:31Hacker News：AI 热帖AI score 64/100

Ember-1

September 24, 20261 story

08:00OpenRouter：Announcements（RSS）AI score 77/100

Is Kimi K3 Open Source? Weights, License, and How to Call It

OpenRouter 撰文说明 Kimi K3 是开放权重而非开源模型，Moonshot AI 以自定义 Kimi K3 License 在 Hugging Face 发布 moonshotai/Kimi-K3。

September 23, 20261 story

00:00Fireworks AI（网页）AI score 66/100

Introducing Ember-1

Fireworks Research 发布基于 Kimi K3 训练的专用模型 Ember-1，通过学习精简不必要的推理，在保持质量的同时减少约 35-50% 的 reasoning token。

September 22, 20261 story

19:00X：Kimi.ai (@Kimi_Moonshot)AI score 69/100

Meet the new Kimi Browser Extension, formerly Kimi WebBridge. From your browser sidebar, you can ch...

Kimi 发布新的 Kimi Browser Extension，前身为 Kimi WebBridge。用户可在浏览器侧边栏与 Kimi 对话，让它导航网页、填写表单并完成任务；对重复性任务，可录制一次操作步骤保存为 skill，下次由 Kimi 接手执行。产品现已在 kimi.com/products/kimi-browser-extension 和 Chrome Web Store 上线。

September 10, 20261 story

00:00Cognition 模型 / Devin 博客（网页）AI score 70/100

Introducing SWE-2: Pushing the Pareto Frontier

Cognition 发布最先进编码模型 SWE-2，基于 2.8T 参数的 Kimi K33 后训练，在 FrontierCode 1.1 Main 取得 50.0%，仅比 Fable 5.1 低不到一分且便宜 64%。

July 27, 20263 stories

08:00Modal 官方工程博客（RSS）AI score 78/100

Kimi K3 by Moonshot now available on Modal

Moonshot 发布 Kimi K3，总参数 2.8T、每 token 激活 16/896 专家，支持 1M token 上下文窗口和原生视觉。Modal 与 Moonshot、vLLM 合作提供 Day 0 支持，通过定制的 DFlash 投机解码模型将智能体负载交互速度从 100 提升到 460 tokens/s，吞吐从 800k 提升到 1.5M TPM/GPU。

08:00vLLM 官方博客（RSS）AI score 75/100

Kimi K3 Is Here: Efficient Day-0 Support on vLLM

vLLM 官方宣布 Day-0 支持 Kimi K3，这是一个 2.8 万亿参数的 MoE 模型（每 token 激活 896 专家中的 16 个），基于 Kimi Delta Attention 与 Attention Residuals，支持 1M token 上下文窗口和原生视觉。

00:00Fireworks AI（网页）AI score 77/100

Kimi K3 on Fireworks: Frontier Intelligence You Can Own

月之暗面开源 Kimi K3，Fireworks Day-0 提供推理与训练支持。模型总参数 2.8T、激活参数 104B、1M token 上下文，Frontend Code Arena 1。

July 26, 20261 story

08:00Together AI 研究与产品博客（RSS）AI score 75/100

Kimi K3 vs GPT-5.6 Sol on DeepSWE: Cost, Coding, and Routing

Together AI 基于 DeepSWE 全部 904 次评分 rollout（113 个任务、各 4 次尝试）对比 Kimi K3 与 GPT-5.6 Sol。

July 18, 20261 story

01:11X：Artificial Analysis (@ArtificialAnlys)AI score 76/100

The frontier has opened up: four frontier launches in eight days - Grok 4.5, GPT-5.6, Muse Spark 1.1...

过去八天内，Grok 4.5、GPT-5.6、Muse Spark 1.1 与 Kimi K3 四款前沿模型相继发布，使 Artificial Analysis Intelligence Index 得分超 50 的实验室从 6 月初的 2 家增至 6 家。

July 14, 20261 story

00:00Moonshot AI：Kimi BlogAI score 81/100

Kimi K3

月之暗面发布 Kimi K3，一个 2.8T 参数的开源模型，采用 Kimi Delta Attention 和 Attention Residuals 架构，支持原生视觉能力与 100 万 token 上下文窗口。

July 8, 20261 story

00:00Cognition 模型 / Devin 博客（网页）AI score 67/100

SWE-1.7: Frontier Intelligence at a Fraction of the Cost

Cognition 发布 SWE-1.7，称其以更低成本达到前沿智能水平。模型基于 Kimi K2.7 底座训练，FrontierCode 1.1 Main 通过率 42.3%，Terminal-Bench 2.1 为 81.5%，已在 Devin（Web、Desktop 和 CLI）中经 Cerebras 以 1000 TPS 提供。

June 12, 20261 story

18:16X：Kimi.ai (@Kimi_Moonshot)AI score 70/100

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced! 🔷 Improved coding & ...

Kimi 发布并开源最新代码模型 Kimi-K2.7-Code。相比 K2.6，其在 Kimi Code Bench v2 上提升 +21.8%，Program Bench 提升 +11.0%，MLS Bench Lite 提升 +31.5%。推理效率改进，推理 token 使用量降低 30%，长时编码任务中指令遵循和端到端成功率均提升。6x 高速模式即将推出，即日起可通过 Kimi API 和 Kimi Code 使用。

May 25, 20261 story

08:00Epoch AI：研究、数据与评测AI score 65/100

Is a compute crunch coming?

Epoch AI 的 Luke Emberson 建模估计，截至 2025 年 Q4 全球 Blackwell GPU（GB200 约 190 万、GB300 约 150 万块）以 Kimi K2.6 类模型可产出每秒 5 亿到 200 亿输出 token，推理容量每年增长约 3.4 倍。

December 12, 20251 story

08:00Thinking Machines Lab：News（网页）AI score 67/100

Tinker: General Availability and Vision Input

Thinking Machines Lab 宣布 Tinker 正式开放（GA），取消候补名单，并新增四项更新：支持微调 Kimi K2 Thinking（万亿参数。

Showing the latest 18 of 34 stories.