6 results for “open weights”
NVIDIA's NemotronLabs VoiceChat 11B is the first open full-duplex speech model with live tool calling, achieving 448 ms turn-taking latency but labeled research-only with no hosted API.
Read more →
Meta launches Muse Code, a terminal AI coding agent powered by Muse Spark 1.2, featuring persistent background agents, parallel worktree isolation, and aggressive pricing that undercuts rivals by up to 21x.
Read more →
Mistral AI releases Shieldstral, a 3B open-weight multimodal moderation model that runs on a single 16GB GPU under Apache 2.0, using prompt-driven policies instead of baked-in taxonomies.
Read more →
Two dozen companies including Meta, Microsoft, Nvidia, and IBM signed an open letter urging US policymakers to protect open-weight AI models, arguing they lower costs, increase competition, and improve security through transparency.
Read more →
Moonshot AI released Kimi K3, a 2.8-trillion-parameter open-weight model that shifts the constraint from compute to memory. The launch is significant, but weights, software support, and deployability still need scrutiny.
Read more →
DeepSeek V4 Flash is an open-weight AI model focused on long context, low-cost inference, and practical coding and agent workflows.
Read more →