🚀 KwaiKAT lands on ZenMux — now 40% OFF!
KAT-Coder-Pro V2.5 is a flagship Agentic Coding model that goes beyond writing code to handling real engineering tasks. 🔥
🛠️ Tackles repo-level issues
🤖 Built for complex agentic workflows
🎯 Strong across coding, tool use, and UI tasks
Now joining the ZenMux Token Economics lineup at 0.6x pricing 👉
zenmux.ai/kuaishou/kat-coder…
#ZenMux #TokenEconomics #KwaiKAT
📢 KAT-Coder-Pro-V2 by @KwaiAICoder — flagship agentic coding model, now live on @ZenMuxAI .
- Native OpenClaw with multi-scaffold support
- Aesthetic web generation from natural language
- Optimized for high-frequency agentic calls
🔗 zenmux.ai/kuaishou/kat-coder…
KwaiKAT retweeted
Them: "How tf do you spend so little on AI coding???"
KAT: "Just $1.7 for the whole first month."
KwaiKAT retweeted
Meet KwaiKAT Coding Plan – powered by KAT‑Coder‑Pro V1
- First month only $1.7 — 10% of Claude Code Pro’s price, with the same usage limits.
- AI model built specifically for agentic coding
- Ideal for vibe and spec coding
Boost your coding now!
KwaiKAT retweeted
Using @KwaiAICoder 's Kat-coder, I built a local MCP server help the planning/spec refinement step for any vibe coded project. Perfect for those IDE that lack a plan mode, but supports MCP. "kat-planner" MCP is born. Project: github.com/spenceriam/kat-pl… #KwaiKATChallenge
KwaiKAT retweeted
Try the latest version on Novita AI. novita.ai/models/model-detai…
KwaiKAT retweeted
Huge congrats to the @KwaiKAT team on KAT‑Coder‑Pro V1’s new results on the @ArtificialAnlys Intelligence Index — now the top non‑reasoning model in their benchmarks.
We’re excited to offer KAT‑Coder‑Pro V1 for free on Atlas Cloud and OpenRouter:
Try it on Atlas Cloud: atlascloud.ai/models/kwaipil…
Try it on OpenRouter: openrouter.ai/kwaipilot/kat-…
KAT-Coder-Pro V1 by KwaiKAT is a non-reasoning model which demonstrates impressive performance across reasoning-type tasks, while using significantly fewer output tokens than peers. It scores 64 on the Artificial Analysis Intelligence Index, the highest of any non-reasoning model.
@KwaiAICoder's KAT-Coder-Pro V1 is a non-reasoning model, meaning the model does not use a distinct chain of thought, or ‘thinking’ trace, before answering. It is a proprietary model and KwaiKAT has not released the weights.
Key Highlights:
➤ 🧠 Intelligence: KAT-Coder-Pro V1 scores 64 on the Artificial Analysis Intelligence Index, and places in line with Grok 4.1 Fast (Reasoning) and Claude Sonnet 4.5 (Thinking). The model displays strength in agentic tool use (89% on Tau2-Telecom), reasoning (33% on HLE) and long context reasoning (74% on AA-LCR). For frontier physics reasoning in CritPT, KAT-Coder-Pro V1 scores 0%
➤📚 Knowledge and Hallucination: KAT-Coder-Pro V1 scores -36 on the Artificial Analysis Omniscience Index, placing it ahead of GPT-5.1 (Non-reasoning) and Gemini 2.5 Flash (Sep, Reasoning). The model scores 18% on accuracy and 66% on hallucination rate
➤⚙️ Token Efficiency: KAT-Coder-Pro V1 is a non-reasoning model and uses fewer tokens than peers of equivalent intelligence. It uses 7.6M output tokens, which is significantly fewer output tokens than peer models of equivalent intelligence such as Grok 4.1 Fast (Reasoning, 71M) and Claude Sonnet 4.5 (Thinking, 42M)
➤💲 Pricing: KAT-Coder-Pro V1 is currently available for free via Streamlake and Novita API endpoints, though pricing may change in the future.
Key Information:
➤ Context Window: 256K tokens
➤ Multi-modal capabilities: Text input and output only
KwaiKAT retweeted
The top non-reasoning coding model is now free in Kilo. @KwaiAICoder's KAT-Coder-Pro V1:
- 73.4% SWE-Bench Verified
- Top 10 on the @ArtificialAnlys Intelligence Index
- Optimized for tool-use, multi-turn interaction, and instruction following via multi-stage RL training
Just select it from the model picker and let us know how it goes 👇
KAT-Coder-Pro V1: HUGE UPDATE!
The latest version of KAT-Coder-Pro V1 hits 64 on the Artificial Analysis Intelligence Index, ranking 10th globally among all models
- Boosted benchmark performance
- Strengthened reasoning capabilities
- Enhanced front-end capabilities
- Elevated tool-use precision
- Reduced hallucination
(1/n)
KwaiKAT retweeted
🤯🤯🤯 Gemini 3 Pro + Live-SWE-agent hits 77.4% on SWE-bench Verified, beating ALL existing models, including Claude 4.5!!
🤖 Live-SWE-agent is the first live software agent that autonomously self-evolves on the fly — and it even outperforms the manually engineered scaffold used by the Gemini 3 Pro team (76.2%)















