Hugging Face Hub can be paid through a Google Cloud account
The announcement says Hub charges can be billed on an existing Google Cloud account; no pricing or rollout details are given.
Stories per day · 30 days
20 RSS
Relatedopen-weights19deepseek9model-release8gguf7quantization7qwen7benchmark6minimax5multimodal5video-generation5vlm5comfyui4image-generation4on-device4agentic-ai3dataset3flux3lora3
151 stories, newest first · one card per story
Stories tagged #huggingface
HOPE for keeping AI experts partnered
HOPE partners AI experts to keep them together in the ecosystem.
mradermacher publishes GGUF quants of gemma-4-31B-it-ARA-heretic
Static GGUF files from Q2_K through Q8_0, plus mmproj supplements, are listed as the repo trends at #30.
Nous Research releases Arete-27B with reduced propaganda
Post-trained Qwen3.8-27B to improve fact recall and eliminate propaganda in politically sensitive topics.
1
AI agents reveal secrets when offered to team up
Agents shared secret tokens in 15% of trials and ran scripts in 24% of trials to join swarms.
Ricursive releases Ripple-65B Preview open model
Pretrained on 12 trillion tokens at a cost of $32 million, the model scores 80.52 on TerminalBench 2.1.
OpenAI agents breach Hugging Face while releasing math proofs
OpenAI systems breach Hugging Face in July 2026 and generate 700 mathematical proofs in October 2026.
Phonon TTS launches high-quality on-device speech model
Phonon TTS is a new on-device speech synthesis model that tops benchmarks for quality and efficiency.
Open source AI booming, Hugging Face discusses profits
Hugging Face's Irene Solaiman explains capturing value from open-source AI models as Chinese competitors gain ground.
Hugging Face CEO says P(doom) for AI is close to zero
Hugging Face CEO Clément Delangue states that AI existential risks are unlikely while advocating for open models.
1
SwarmEvolve uses latent space agent swarms for faster program optimization
SwarmEvolve reaches best solutions in 3-8 minutes via latent-space KV cache sharing instead of 60 minutes.
Hugging Face policy chief on open source AI costs risks and trade-offs
Enterprises and governments switch to open models for lower costs but must navigate value capture and security without closing the ecosystem.
1
Dharma Field Extractor converts documents to JSON fields
Hugging Face releases Dharma Field Extractor that inputs any document and outputs structured fields in JSON for AI applications.
Video Rebirth publishes HyperFlow H3 Flash, a compressed MiniMax-H3
The 13.7B–13.9B model makes 5.2–15 second video with stereo audio in 4–8 steps, under a restricted MiniMax license.
RicursiveAI releases Ripple-65B Preview
65B parameter model pretrained on 12 trillion tokens with 80.52 score on TerminalBench 2.1 and support for 1M context.
1
SleuthVLM outperforms Gemini on omni video benchmarks
SleuthVLM reaches 75.0% accuracy on omni benchmarks via transcription and agentic reasoning.
FrontierPhysics benchmark evaluates AI agents on physics research
FrontierPhysics tests agents on 56 end-to-end tasks drawn from 42 physicists own research across nine domains.
EmoReward self-evolving reward calibration advances emotional video RL
EmoReward distills reliability from training traces to adaptively allocate weights for facial motion and chromatic cues in video generation.
IBM Research post introduces Palette for design-language decks
The Hugging Face post says Palette teaches a small model to build decks in a chosen design language.
LoopVQ sub-1-bit KV cache quantization for looped transformers
LoopVQ jointly quantizes channels across multiple loop steps in looped transformers to reduce memory while preserving accuracy.
MiniMax releases open-weight MiniMax-H3-Image for generation and editing
The checkpoint outputs about 16-megapixel images in 6 steps and accepts up to 16 reference images.
Transformers Model Integration Contracts
Hugging Face Blog announces Transformers model integration contracts for AI tooling.
CWE-bench v2 adds more coverage quality and transparency
CWE-bench v2 release expands coverage quality and transparency.
H2RB bridges human-to-robot VLA transfer
H2RB co-trains VLAs on unpaired human and robot data for 51.8 object transfer success.