llama.cpp adds support for decision models that score options instead of writing text

The llama.cpp server now supports decision models through a new endpoint. These models read an input once and return a probability for each option you give them.

Ai2 open-sources AstaBrief, a small model for writing cited research reports

The Allen Institute for AI released AstaBrief 8B, which turns a research question and retrieved literature excerpts into a cited report. It powers the Fast mode of report generation in Ai2's Asta platform.

Hugging Face launches an open, multilingual text-to-speech leaderboard

Hugging Face and collaborators launched the Open TTS Leaderboard to evaluate open-source text-to-speech and voice-cloning models across languages. The authors say open models are underrepresented on arena-style rankings.

H Company releases Holo4 models for agents that operate software

Paris-based H Company introduced Holo4, agentic models in 27B dense and 35B-A3B mixture-of-experts sizes, plus an updated Holotron4 Nano. The models work through GUIs, code, MCP and APIs.

Hugging Face Hub adds a home for reinforcement learning environments

Hugging Face added an RL Environments filter to the Hub so agent training and evaluation environments can be shared as dataset repos. It works with frameworks including Harbor, Verifiers, OpenEnv and NVIDIA NeMo Gym.

ESPnet releases YODAS v3, a 1.1-million-hour open speech dataset

The ESPnet team published YODAS v3, an open dataset of about 1.1 million hours of real-world speech in more than 100 languages, with 48kHz audio and timestamped transcripts.