MoonshotAI Kimi-K2.5 azure/kimi-k2.5 Input: text Output: text RuntimeWire coverage OpenAI reports up to 3.6x lower Jalapeno latency in engineering testsOpenAI says Jalapeno beats Nvidia Blackwell on inference speed per wattResearchers evolved AI "mind viruses." The antivirus was one paragraphBaseten ships Rust tokenizer it says is 18x faster for Kimi K3Kimi K3 produces 19 exploits in 90 minutes for the latest Redis releaseSpaceXAI releases Grok 4.5 with Cursor-trained coding push View on OpenRouter. Model data sourced from OpenRouter.