Загрузка видео...
Не удалось загрузить видео
After testing Kimi.ai K2 for a few hours, My overall take: - Performance between Claude 3.5 & Claude 4 (Just my vibe eval!) - The UI generation seems great - But the cost is only 20% of Claude 3.5 - So good enough for most coding agent with a... show more
80,871 просмотров • 1 год назад •via X (Twitter)
Комментарии: 11
If you are interested, I will share Claude Code best practices & how to integrate Kimi K2 in the next @aibuilderclub_ weekly session ----- btw you can get Moonshot API keys here: Or use @OpenRouterAI { "LOG": false, "Providers": [ { "name": "openrouter", "api_base_url": " "api_key": "sk-or-v1", "models": [ "moonshotai/kimi-k2:free" ], "transformer": { "use": [ "openrouter" ] } }, "Router": { "default": "openrouter,moonshotai/kimi-k2:free" } }
@Kimi_Moonshot I can select Kimi K2 with @OpenCode_AI using @OpenRouterAI. I haven't tested that yet but that seems like the more "intended" strategy.
Scan any documents, convert images into text, PDF files, etc. 👍
@Kimi_Moonshot That cost difference is the real story for agents. Token usage adds up fast, so 20% of 3.5 is massive. Good to hear it holds up performance-wise for those tasks.
@Kimi_Moonshot I also tried it and it really well
@Kimi_Moonshot alright, waiting for groq/cerebras to ship cheap inference of this model
@Kimi_Moonshot Can we use both Claude and Komi at the same time?
@Kimi_Moonshot yes you can
@Kimi_Moonshot How about in Cline?
@Kimi_Moonshot I fucking love china man
@Kimi_Moonshot Interested
![5 GitHub repos with ⭐50K+ combined stars that show how far people are pushing Claude Code beyond a single coding agent 1) instructkr/claude-code 2.3K⭐ -> a research snapshot of Claude Code’s source with the internals exposed for inspection. [ ] 2) codeaashu/claude-code 3.2K⭐ -> another reconstructed Claude Code codebase built for studying how the agent works under the hood, including tools, prompts, MCP. [ ] 3) Donchitos/Claude-Code-Game-Studios 20K+⭐ -> turns one Claude Code session into a game development studio with 49 specialized agents, 73 skills, 12 hooks, 11 rules and 41 templates. [ ] 4) VoltAgent/awesome-claude-code-subagents 24K+⭐ -> 154 specialized Claude Code subagents split across 10 categories. [ ] 5) tanbiralam/claude-code 2.2K⭐ -> a research-oriented Claude Code source reconstruction for inspecting the terminal agent architecture, plugins, tools and internal implementation. [ ] Save these if you want to understand what people are actually building around Claude Code instead of just using the default CLI.](https://image.24vids.com/tw-2090119203810320462/amplify_video_thumb/2090119067331956736/img/tm1pipu-4wl7-_AJ.jpg)
