Prime Intellect Buys Moondream to Double Vision AI Throughput
The neocloud is folding Moondream's open vision models, Photon inference engine, and megakernel compiler into its stack to chase cheaper multimodal inference.
20 stories
The neocloud is folding Moondream's open vision models, Photon inference engine, and megakernel compiler into its stack to chase cheaper multimodal inference.
A new discrete-diffusion approach builds phylogenetic trees step by step, outperforming Phyloformer 2 and matching MCMC under realistic evolutionary models.
Cognition's Devin now plugs into Pylon so customer bug reports flow straight into AI coding sessions, with $15,000 in credits for new teams.
Liquid AI's new 24B MoE activates just 2B parameters per token, hits 184 tok/s on a MacBook, and targets fully local agents.
A new RLVR training method called Proximal KL swaps PPO-style clipping for bounded future-KL credit, beating GRPO by up to 2.58% on math benchmarks.
Arena.ai reworked its bash evaluation to credit first-try command-line reliability, reshuffling how agentic models rank across real user sessions.
Anthropic replaced git with Scion, an in-house version control system built for agents, after daily merges grew from 2,500 to over 30,000.
A new leaderboard pits purpose-built decision models against LLMs on classification, tool calling, knowledge, and reasoning, with cost and latency side by side.
An open-source Mac menu bar app uses Google's EmbeddingGemma 2 to deliver fully local semantic search across photos, PDFs, audio, video and code.
Cognition added Google's new frontier model to Devin Desktop and CLI, where it tops Gemini scores on FrontierCode at 60.5% with a 75% launch discount.
Google's new frontier model tops the Vals Index at 68.9%, ships with a 1M token window, and launches at $2/$10 per million tokens.
A 4-bit MLX conversion of Qwen3.8-27B squeezes a native multimodal dense model into roughly 16 GB, making it practical on Apple Silicon Macs.
Fish Audio's new 35B mixture-of-experts text-to-speech model ships open weights, with plain-language direction, mid-sentence voice shifts, and word-level audio inpainting.
A new family of video models skips autoregressive text generation to score time spans and answer options directly, hitting over 100x QA speedups.
AutoTrust AI's open-weights JEV-35B turns a Qwen3.5 MoE into a System 1 decision model that picks among 256 options in 204 ms.
A researcher converted Qwen3.8-27B into a masked diffusion language model that generates tokens in parallel instead of left to right, with open weights and code.
Sarvam's new Bulbul V4 Flash delivers sub-250ms Indian-language speech synthesis at a third less cost, with 83 voices across 11 languages.
A trending open-source toolkit brings reproducible release gates, A/B ship decisions, and offline-verifiable records to real-time ML decision systems.
Gen-Verse's open-source framework pairs world-model predictions with real feedback, pushing a single recursive loop across code, GUI, chemistry and biology.
Alibaba's Qwen team shipped an 8-step accelerated version of their 7B open-weight image model, cutting denoising work by 5x while keeping 2K output quality.