Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Chinese AI company 01AI_Yi has released two new code language models, Yi-Coder 1.5B and Yi-Coder 9B, under the Apache 2.0 license. These models, which support 52 major programming languages, offer state-of-the-art performance in code editing, completion, debugging, and math reasoning. The models excel in long-context understanding with a maximum context length of 128K tokens. Yi-Coder 9B, continuously pretrained with 2.4 trillion tokens, outperforms larger models such as DeepSeek-Coder 33B on LiveCodeBench, achieving a 23.4% pass rate. The models are available in both Chat and Base versions.
DeepSeek-V2.5 🚀 an OPEN model combining general and coding capabilities just released by Chinese AI unicorn @deepseek_ai . https://t.co/yE2AqSuwr7 ✨ Combine DeepSeek-V2-Chat & DeepSeek-Coder-V2 ✨ Enhanced writing, instruction-following and human preference alignment
Big! @deepseek_ai-V2.5 released! 🚀 A “merge” of DeepSeek-V2-0628 and DeepSeek-Coder-V2-0724 improved across benchmarks with native Function Calling and JSON Mode! 👀 📊 236B parameters, requires 80GB*8 GPUs for BF16 inference 🔢 Improved benchmarks: AlpacaEval 2.0: 50.5 (up… https://t.co/u50dyZ8vxn
DeepSeek 2.5 is out! A powerful MOE with 238B params with 160 experts and 16B active params 👀Chat and code capabilities 🔥Function calling 💻JSON output and FIM completion 📏128k context length They also dropped DeepSeek-Coder-V2 Check it out! https://t.co/pD7OQSpVy4
