UGeeker

[Elliot Arledge @elliotarledge]

https://t.co/Hzx85f3xu3

[熊师傅 weight decay 了吗 @bigeagle_xd]

https://t.co/MWUVsw1Ypx

K3 发布前两天内部 vibe 出来的 win xp 模拟器，功能相当完整，有“千千静听”可以听歌，有“QQ2005”可以聊天，朋友们会在你的QQ空间里聊超女，新浪首页在报道神州6号……

在内部传开的那个下午，站起身来，发现一半的同事都在用 XP，瞬间有一些恍惚，此情此景，���佛回到了学校的微机课，又像是当年的网吧。我回想起 20 年前中考后的夏天，每天在家听着《宁夏》和《突然很安静》，在不舍得开空调的炎热的下午，在家里的老电脑上刷冒险岛……

20 年弹指一挥间，感谢 K3，能在这个剧变的时代，让我怀念一下年轻的自己。

[熊师傅 weight decay 了吗 @bigeagle_xd]

the fifth should be replaced with “不要比烂”.

think not what you've done better than someone else, but what other labs have done better than you.

oai and a\ are still leading the real frontier, glm can match opus with a <1T model, dpsk has incredibly high inference efficiency, we have very much work to do!

[Anvita Flow @AnvitaFlow]

The"Tool" era of AI is ending. The "Agentic" era is here—but something is missing. 🧩
Individual agents are silos. Fragmented logic, rented intelligence, and isolated souls.
We’ve been building brains, but we forgot the Flow.
Introducing Anvita Flow: The Onchain Agentic Collaboration Network. 🌊
Over the next 4 days, we are going to ask 4 questions that will redefine how you see AGI, ownership, and human connection.
Are you ready to stop the silos and start the flow?
#AnvitaFlow #AgenticEconomy #Web3 #AI #RethinkingAGI

[熊师傅 weight decay 了吗 @bigeagle_xd]

还是不够好，意难平，下次一定

[Kimi.ai @Kimi_Moonshot]

Kimi K2-0905 update 🚀
- Enhanced coding capabilities, esp. front-end & tool-calling
- Context length extended to 256k tokens
- Improved integration with various agent scaffolds (e.g., Claude Code, Roo Code, etc)

🔗 Weights & code: https://t.co/SsFKTnWslD
💬 Chat with new Kimi K2 on: https://t.co/2bLWEHF6az
⚡️ For 60–100 TPS + guaranteed 100% tool-call accuracy, try our turbo API: https://t.co/EOZkbOwCN4

[LMSYS Org @lmsysorg]

🚨SGLang Summer Fest Bonus Drop🚨
Proud to share a joint effort from Mooncake by
@Kimi_Moonshot, @Oracle , and SGLang: Kimi K2 trillion-scale deployment—running on 128 H200 GPUs sponsored by @NVIDIAAIDev DGX Cloud. OME + SGLang = MoE inference at production scale.👇

[elvis @omarsar0]

Agent Leaderboard v2 is here!

&gt; GPT-4.1 leads
&gt; Gemini-2.5-flash excels at tool selection
&gt; Kimi K2 is the top open-source model
&gt; Grok 4 falls short
&gt; Reasoning models lag behind
&gt; No single model dominates all domains

More below:

[Arena.ai @arena]

🚨 BREAKING: @Kimi_Moonshot’s Kimi-K2 is now the #1 open model in the Arena!

With over 3K community votes, it ranks #5 overall, overtaking DeepSeek as the top open model.

Huge congrats to the Moonshot team on this impressive milestone! The leaderboard now features 7 different providers in the top 15 - the most competitive it’s ever been.

More insights in the thread 🧵

[DeepInfra @DeepInfra]

Up to 100 tps Moonshot AI's Kimi K2, as always at the best price of $0.55/$2.20. Zero data retention, generous rate limits.

[UGeeker @UltimateGeeker]

Finally! somebody found this !!!!

The Agentic loop &amp; Coding are REAL "THINKING"

[Dan McAteer @daniel_mac8]

🔥 Kimi K2 is not a reasoning model but you can use it like one

🎯 here's how:

1. give it access to the 'Sequential Thinking' MCP (link below)

2. put it in an agent loop

3. tell it to think sequentially before answering

it's so cheap that it won't cost you that much.
use a fast inference provider like Groq and it'll be super fast too

💥 BOOM. you've got a reasoning-like model.
(yes I understand that this is not quite the same as an internal CoT but regardless should make it smarter)

[UGeeker @UltimateGeeker]

If you use the official API, there are tons of input tokens will hit cache when using coding agents. My rough estimation is like around 85% tokens are hit in typical agentic usage, that will cut the cost dramatically compared to third party providers for Kimi K2.

[OpenHands @OpenHandsDev]

Kimi-K2 is definitely the first strong open-weight competitor to Claude Sonnet.

65.4% on SWE-Bench Verified in OpenHands, just 2.6 points shy of Claude Sonnet 4 with the same 100 iterations.

API cost is 4x cheaper, together with all benefits of open weights.

[Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex]

Kimi-K2-reasoner is going to be so lit

[UGeeker @UltimateGeeker]

@oneitonitram @Kimi_Moonshot We are cooking a much larger one

[UGeeker @UltimateGeeker]

@bitdeep_ @Kimi_Moonshot we won’t

[OpenRouter @OpenRouter]

Grok 4 and Kimi K2 competing on top of the Trending models charts

[LinearUncle @LinearUncle]

opencode是目前开源项目中最好的coding agent.
你看，它也有todo list，sub agent etc.

天下苦anthropic封号久已，kimi k2的诞生，组合openCode，让vibe coding事业有了更多的选择。

[UGeeker @UltimateGeeker]

Kimi K2 will thank you

[Elon Musk @elonmusk]

Grōk

[Elon Musk @elonmusk]

Grōk

[Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex]

Kimi K2 at 185 t/s (or even higher, nearly 220 in my short tests) is probably the best use of Groq to date, and can make K2 immediately more compelling than Sonnet 4. Impressive that they've managed to fit this 1T monster on their chips.