CODEX / SIGNAL STUDIO
Perplexity Developers has made pplx-decider-v1.1-27b available as an updated open-weights multimodal decision model. Priced at $0.02 per million input tokens, the model costs half as much as v1 and ranks highest on the Hugging Face Decision Index 0.3 benchmark according to the team.
Google detailed operational specifications for EmbeddingGemma 2, stating the 740M model uses between ~191MB and 567MB of active RAM, expands its context window to 8K, and processes up to 5.5 minutes of audio, 29 images, or 58 video frames per pass.
EmbeddingGemma 2 runs multimodal workloads entirely offline with zero server calls. The model features a modular design that lets developers drop unused vision or audio components to save memory, offers flexible dimension sizes that reduce local database storage by up to 6x, and pairs with Gemma 4 for efficient on-device RAG pipelines.
GeoGuess Bench evaluates AI models on 210 worldwide geolocation photo tasks. Claude Opus 5.5 ranked first overall, while open models delivered competitive accuracy at lower inference costs.
Mistral AI announced Mistral Large 4, a 1-trillion parameter natively multimodal model with 49 billion active parameters. The model is available now via API and Mistral Cloud infrastructure, ahead of an open weights release planned for late October.
