NadiAIGlobal AI news. Concise. Clear. Easy to follow.
Z.ai has confirmed it is the lab behind the previously mysterious Ox Alpha model, which has been topping benchmarks and leaderboards. The company says the model's weights will be released soon, according to TechCrunch AI.
Benchmarking on SemiAnalysis’ InferenceX suite showed OpenAI’s Jalapeño chip delivered higher tokens per user and greater throughput per kilowatt than the current state-of-the-art. TechCrunch AI reported these test results without independent verification in the article metadata.
Nous Research melancarkan NousCoder-14B, model penjana kod sumber terbuka yang dilatih dalam empat hari menggunakan 48 GPU Nvidia B200 dan mencapai 67.87% pada LiveCodeBench v6. Syarikat itu turut menerbitkan berat model, rangkaian latihan Atropos dan alat pengesahan, serta memberi amaran bahawa set data verifikasi kompetitif (≈24,000 soalan) mungkin hampir mencapai had kualiti tersedia.
