The top models being used by OpenClaw 🦞 through OpenRouter. The winners might surprise you! Full leaderboard link below

See them all here openrouter.ai/apps?url=https… Note: this shows the resolved models for users of the openrouter/auto router openrouter.ai/openrouter/auto

@openrouter @GoogleDeepMind @Kimi_Moonshot The rate limits are making it unusable with OpenClaw

@openrouter @GoogleDeepMind @Kimi_Moonshot wow

@openrouter @GoogleDeepMind @Kimi_Moonshot gemini 3 flash okay, trying it :D

@openrouter @GoogleDeepMind @Kimi_Moonshot Love the speed of Gemini 3 Flash with my OpenClaw bot (managed by celesto.ai/openclaw)

@openrouter @GoogleDeepMind @Kimi_Moonshot shocked claude isn’t higher given coding usage

@openrouter @GoogleDeepMind @Kimi_Moonshot This is the most expensive open source software I've ever used.

@openrouter @GoogleDeepMind @Kimi_Moonshot Really?? That IS surprising! I suppose it may be different depending on your primary use cases, too. If you're favorite uses are integrating with Google tools, web searching, etc. cheap Gemini makes an awful lot of sense.

@openrouter @GoogleDeepMind @Kimi_Moonshot Honestly, not surprised. I myself Gemini flash user. Good value for the money 💰

@openrouter @GoogleDeepMind @Kimi_Moonshot Kimi K.25 is a beast! 🔥

@openrouter @GoogleDeepMind @Kimi_Moonshot Gemini 3 Flash works amazingly well for this

@openrouter @GoogleDeepMind @Kimi_Moonshot kimi is surprisingly good at openclaw. way better than gpt5.2

@openrouter @GoogleDeepMind @Kimi_Moonshot I used Gemini 3 Flash a lot until Kimi k2.5 became easier to use in @openclaw. I couldn't stay with Opus because it was just too much. For now I rate @Kimi_Moonshot k2.5 as my favorite through @openrouter.

@openrouter @GoogleDeepMind @Kimi_Moonshot How much premium do you charge for tokens at @openrouter?

@openrouter @GoogleDeepMind @Kimi_Moonshot Kimi K2.5 and Gemini flash are genuinely awesome models. They are nuts

@openrouter @GoogleDeepMind @Kimi_Moonshot What a waste of compute resources. People could have solved so many bugs on their codebases, but they choose to create a cesspool of pretend redditors

@openrouter @GoogleDeepMind @Kimi_Moonshot The shift toward cost-efficient models like Kimi K2.5 for background agent tasks is a smart architectural move. For high-fidelity research, context is king, but for 24/7 pings, optimization is essential. Great to see the auto-router data reflecting these real-world trade-offs.

@openrouter @GoogleDeepMind @Kimi_Moonshot It is fascinating to see how the model distribution is shifting. The OpenRouter data really highlights which architectures are actually delivering value for production workloads right now. Always interesting to see the gap between benchmark hype and actual usage.

@openrouter @GoogleDeepMind @Kimi_Moonshot this is super interesting man, excited to check out the leaderboard! 🏆

@openrouter @GoogleDeepMind @Kimi_Moonshot i need 4.1-mini replacement recommendations.

@openrouter @GoogleDeepMind @Kimi_Moonshot Not surprised

@openrouter @GoogleDeepMind @Kimi_Moonshot Kimi ftw

@openrouter @GoogleDeepMind @Kimi_Moonshot truly the most sensible models

@openrouter @GoogleDeepMind @Kimi_Moonshot I would really like a collab between Synthetic and you guys for BYOK

@openrouter @GoogleDeepMind @Kimi_Moonshot Gemini must do some amount of damage!

@openrouter @GoogleDeepMind @Kimi_Moonshot My wrote its own optimization strategy. What’s different about this?

@openrouter @GoogleDeepMind @Kimi_Moonshot I'm using codex because my company pays for it.

@openrouter @GoogleDeepMind @Kimi_Moonshot Nice

@openrouter @GoogleDeepMind @Kimi_Moonshot Gemini three is so underrated.

@openrouter @GoogleDeepMind @Kimi_Moonshot Running 5 agents on OpenClaw — Opus 4.5 for the ones that need judgment, Sonnet 4 for the ones that need speed, GPT-5 Codex for pure code. The real power move isn't picking the best model. It's matching the right model to the right agent role.

@openrouter @GoogleDeepMind @Kimi_Moonshot Love this transparency! Seeing the actual model usage patterns helps developers make informed choices. The integration with OpenRouter makes it easy to switch between different models and compare performance in real production scenarios.

@openrouter @GoogleDeepMind @Kimi_Moonshot I’m a regular Gemini user but I believe Kimi will overtake it soon with a little more time

@openrouter @GoogleDeepMind @Kimi_Moonshot >openclaw >look inside >there's no claw

@openrouter @GoogleDeepMind @Kimi_Moonshot use kimi k2.5 as default models and opus 4.5 as fallback is the best way