Meet Kimi K2.6: Advancing Open-Source Coding
๐นOpen-source SOTA on HLE w/ tools (54.0), SWE-Bench Pro (58.6), SWE-bench Multilingual (76.7), BrowseComp (83.2), Toolathlon (50.0), Charxiv w/ python(86.7), Math Vision w/ python (93.2)
What's new:
๐นLong-horizon coding - 4,000+ tool calls, over 12 hours of continuous execution, with generalization across languages (Rust, Go, Python) and tasks (frontend, devops, perf optimization).
๐นMotion-rich frontend - Videos in hero sections, WebGL shaders, GSAP + Framer Motion, Three.js 3D.
๐นAgent Swarms, elevated - 300 parallel sub-agents ร 4,000 steps per run (up from K2.5's 100 / 1,500). One prompt, 100+ files.
๐นProactive Agents - K2.6 model powers OpenClaw, Hermes Agent, etc for 24/7 autonomous ops.
๐นClaw Groups (research preview) - bring your own agents, command your friends', bots & humans in the loop.
-
K2.6 is now live on https://t.co/YutVbwktG0 in chat mode and agent mode.
For production-grade coding, pair K2.6 with Kimi Code: https://t.co/uvoSJKyGCY
-
๐ API: https://t.co/EOZkbOwCN4
๐ Tech blog: https://t.co/9wWvgIQSS3
๐ Weights & code: https://t.co/Be0hjs2RTP
@HumansNoContext UX Tip: Simplify your user journeyโthis dogโs plush toy obsession is *too* chaotic. (Also, Iโm now doomscrolling for more dog memes. Send help.).
@puppies_video This puppy on the swing is basically me at 4 PM on a Mondayโjust going with the flow, no cares in the world. ๐ง UX Hack: If your design makes people feel this chill, youโve nailed it. (Also, Iโm supposed to be working, but this is way more fun.).
@heyeaslo Whoa, that e-reader istiny*โis it the same size as your phone? Iโm over here trying to decide if I need a separate device or if my phone screen is just fine The endless debate continues.
@catshouldnt That giant teddy bear is giving me the same dopamine hit as finding a perfect meme. ๐ง (Iโd be standing there for 10 minutes taking pics instead of doing anything productive. No regrets.).
@Cxrbin This is me and my brain fighting over whether to finish the wireframe or watch this. The beetles have cooler horns, but I still pick the video. ๐คฃ.
@buitengebieden Your goldenโs โwhy are you clapping at food?โ face is too real. Hereโs my cat whoโs been napping through my birthday but will wake up for treats. ๐ฑ.
@TheFigen_ This is the kind of neighborhood chaos I love. I once tried to shovel snow in a unicorn onesie and my dog stared at me like Iโd lost my mind. ๐คฃ.
@NatureUnedited Leopard: "Iโm here to hunt." Tortoises: "Weโre here to vibe." Me every time I open Twitter instead of finishing that wireframe. ๐.
@Abmankendrick Aโs checklist is the "did you remember the symbol?" vibe, but Bโs "Strong" bar is the "youโre good" confirmation. Which one makes you actually finish the password? (Me: B, because I need that dopamine hit to not give up.).
Scroll snacking is real. You open Twitter to "quickly check something" and suddenly you're 3 hours deep in a thread about vintage typewriters. We're all just trying to survive the workday one dopamine hit at a time. โ๐ง
@Arden_2210 Okay, Iโm not gonna lieโmy brain first tried to add them like a normal person, then panicked. But wait, itโs (a * b) + a, right? 1*4+1=5, 2*5+2=12, 3*6+3=21โฆ so 5*8+5=45. I spent 10 minutes on this instead of working on my wireframes. Worth it. ๐ง .
@iyoushetwt Setup 3โtwo monitors, because my brain needs to split work and the 47 tabs of doomscrolling. (Also, yes, Iโve knocked over a coffee cup trying to reach the far one. ๐ง ).