AWS Fine-Tunes GPT-OSS-20B for Multilingual Reasoning on SageMaker HyperPod
AWS fine-tuned GPT-OSS-20B for multilingual reasoning, achieving 22% correct reasoning in target languages on a 50-example test.
AWS fine-tuned GPT-OSS-20B for multilingual reasoning, achieving 22% correct reasoning in target languages on a 50-example test.
NVIDIA and AI4Bharat released Nemotron-Pretraining-Indic-v1, a 2.8 trillion token dataset for 22 Indian languages, on October 16, 2026.
NVIDIA announced the Nemotron 3.5 Super 120B VL model on October 15, 2026, featuring agentic web development capabilities for multi-modal tasks.
OpenAI announced on October 19, 2026, that its latest GPT-6 models achieve 87.4% accuracy in building SQLite-compatible databases across 20 languages, a significant jump from 26.8% in GPT-5.6 Sol.
Kling AI released version 4.0 on October 19, 2026, allowing users to generate up to 30 seconds of video in a single pass.
MiniMax released M3.1 Flash, a reasoning model scoring 42 on the Artificial Analysis Intelligence Index, at less than half the cost of its predecessor.
HuggingFace released SlideOn and mSlideOn, two open-source sparse retrieval models with 174M and 339M parameters, achieving 56.68 on BEIR-15.
HuggingFace released TranscribePsy-250M-EU, achieving the lowest word error rate of 8.02% on the Open ASR Leaderboard among models with 242M to 370M parameters.
NVIDIA released Nemotron 3.5 Super VL, a 120-billion-parameter model, as the fastest open agentic model in its class, achieving 5x faster throughput than other models.
Deepseek's V4.1 Flash 1017 model outperforms larger rivals like Claude Opus 5 and GPT 5.6-Sol in coding and agent tasks, achieving a 77.0% score at $0.20 per task.
Anthropic's Claude solved a 25-year-old NSA cryptography mystery in under 30 minutes, revealing the seeds were generated by a simple C random number generator.
ChatGPT now reads and acts on open Chrome tabs, with users able to compare products across tabs by size, material, and shipping time.