Tag
154 articles
Elorian's new AI model Glimpse outperforms GPT-6 Astra in scientific chart reasoning and video captioning, marking a significant advancement in visual AI capabilities.
Learn how to set up and use the Step 5 Preview open weights model from StepFun, a 600B parameter sparse Mixture-of-Experts model with 27B active parameters and 1M token context window.
DeepSeek has released DeepSeek-V4.1-Flash-1017, a performance-enhanced model scoring 77.0 on DeepSWE, alongside a new Search API for developers.
Deepseek's new V4.1-Flash-1017 model outperforms larger rivals in agentic tasks while significantly reducing per-task costs. The company also launched new Search and Fetch APIs to support broader developer integration.
Kyutai, in collaboration with Inria and École des Ponts, has introduced REKI, a 2.4 billion parameter vision-language-action model that brings efficient memory to robot manipulation. The model outperforms larger systems on benchmarks and is built entirely on public data.
Learn how Beam, a new open-weight AI model from Reflection, uses smart techniques to match top AI performance while using much less computing power.
Aleph Alpha has released Kolibri, a German-English mixture-of-experts model with 78 billion parameters, emphasizing European AI sovereignty and open access.
Reflection introduces Beam, an open-weight AI model that rivals Chinese AI systems at a lower compute cost, targeting enterprises and sovereign nations seeking localized AI solutions.
This article explains the concept of large language models using Alibaba's Qwen as an example, showing how models grow in complexity and capability over time.
OpenAI introduces GPT-6.1 Sol, a high-performance AI model delivering near-Astra intelligence at one-fifth the cost of Astra's API pricing.
Microsoft AI has released MAI-Transcribe-2-Streaming, a real-time speech-to-text model that ranks #1 on the Artificial Analysis AA-WER Streaming benchmark. It delivers 2.5% WER with sub-second processing times across 60 languages.
Learn how Alibaba's new Qwen-Audio-3.1-Realtime AI model enables natural, full-duplex voice conversations by listening, thinking, and deciding when to speak.
