Open-Source DiscoveryRSI-27B Beats GPT-6 at Predicting Future Scientific Research
A 27B open model learns to generate scientific hypotheses recursively, matching real published research at nearly triple the rate of GPT-6 and Claude Opus 5.
20 stories
A 27B open model learns to generate scientific hypotheses recursively, matching real published research at nearly triple the rate of GPT-6 and Claude Opus 5.
Anthropic's STEM Fellows gave Claude Mythos 5 plain-language briefs for eight quantum circuits, and it beat the best published designs, sometimes by over 10x.
Zephon plugs into TorchTitan to deliver identical global batches across GPU counts, so a job resumed on different hardware trains on the exact same data.
Factory makes Delegations generally available, letting teams hand coding tasks to its Droid agent directly from Slack, Linear, Jira, and Microsoft Teams.
Kyutai's REKI is a 2.4B vision-language-action model trained only on public data, matching models twice its size with a cached memory trick.
Fish Audio's new 35B mixture-of-experts TTS model takes bracketed stage directions, tops open-weight leaderboards, and ships free under a revenue cap.
Your agent harness is an architectural decision that directly dictates task success and API costs. Here is how to match it to your workload.
Ant Group's InclusionAI opens weights for a 560B hybrid-linear MoE model with 25B active parameters, 1M context, and MIT licensing.
Reflection AI opens the weights of Beam, a 501B Mixture-of-Experts model, alongside a 40+ partner ecosystem and a full agent toolchain.
Comfy-Org repackaged the OpenMOSS 5B diffusion transformer for ComfyUI, bringing JSON layout prompts and strong text rendering to local workflows under Apache 2.0.
A new training recipe called Active Indexing teaches LLMs to cite the documents they learned from, no retriever required at inference time.
Tencent Hunyuan open-sources a 30B memory writer model, a dual-system memory SDK, and MemHub to unify memory across AI tools.
A community-fine-tuned Turkish speech recognition model built on Meta's XLS-R backbone has quietly crossed 893K downloads on Hugging Face.
Unitree's new UniUni software coordinates fleets of its humanoid and quadruped robots for choreographed formations, live monitoring, and emergency stops.
A causal inference study on Scottish households shows income, not driveways, is the real gatekeeper to EV adoption, with standard models overstating the parking effect.
NVIDIA's Vera Rubin NVL72 posts up to 5.7x the throughput per megawatt of GB300 on realistic agentic coding benchmarks.
Meta's open-source audio generation library bundles MusicGen, AudioGen, EnCodec, MAGNeT, JASCO and AudioSeal into one PyTorch toolkit for text-to-music research.
An open-source desktop canvas lets creators chain prompts, images, and video models into reusable AI storyboarding workflows, with assets stored locally.
A new conformal prediction method tightens safety margins for adaptive controllers, letting a learning quadcopter squeeze through gaps a worst-case bound would block.
Sakana AI's QD-Rec evolves a diverse archive of item plus reason pairs instead of a single ranked list, aimed at high-stakes decisions.