AI competition is heating up: Code Arena WebDev's "Best by AI Lab" one-year trend
AI competition is heating up: Code Arena WebDev's "Best by AI Lab" one-year trend: The flagships of China and the United States compete on a weekly basis, and the score difference has changed from a gap to a dozen or twenty points!
I trained an adapter on Qwen3.8-27B that speed up 1.71x prefill (ref kis LLKVApprox took the learning from DeepSeek-V4.1-Flash CED to qwen series)
3,396 tok/s prefill (1.71x @ 6,603 tok, 69.8% greedy match vs baseline untrained projector) - 147 tok/s decode single stream
Ekko Studio 1.0.3 is now available on Google Play.
This release introduces our new Ekko Studio branding, Agent management, Grok and DeepSeek Harness support, improved Coding Agent configuration, and support for OpenCode and its free models. We’ve also added GitHub sign-in,
Ekko Studio 1.0.3 is now available on Google Play.
This update officially enables the Ekko Studio brand, adds Agent management, Grok and DeepSeek Harness support, improves Coding Agent configuration, and supports OpenCode and its free models. Add GitHub
You have reached the end of the archive
All of deepseek