NVIDIA's Parakeet v3 Now Runs Without Python or a GPU
A community port by the handy-computer team converts NVIDIA's multilingual Parakeet-TDT-0.6B-v3 into GGUF files for transcribe.cpp, enabling local CPU and GPU inference.
20 stories
A community port by the handy-computer team converts NVIDIA's multilingual Parakeet-TDT-0.6B-v3 into GGUF files for transcribe.cpp, enabling local CPU and GPU inference.
Meta built a scheduler that scavenges idle GPU and CPU slivers across its inference fleet to improve ads ranking without buying new hardware.
AssemblyAI's first text-to-speech model streams audio in 47 ms, prices at $15 per million characters, and ships with 4,000 voices plus instant cloning.
Zed 1.26 ships expandable variable inspection in debugger hovers, built-in Pyrefly for Python, and DWARF debugging for WebAssembly sources in the JS debugger.
NVIDIA's open-weight 30B mixture-of-experts speech model handles transcription, diarization, translation and audio Q&A in 40 languages from one checkpoint.
Mistral launches an OpenAI-compatible Responses API in public preview and adds DeepSeek V4.1 Flash and GLM 5.3 Flash to its inference lineup.
A new survey reframes foundation model adaptation for medical imaging as a risk-control problem, mapping five mechanisms to clinical deployment trade-offs.
Vals AI tested 25 frontier models on 262 expert-written U.S. immigration questions and found even the best model leaves out one in four essentials.
Apex Intelligence's first open model fine-tunes Qwen3.6-35B-A3B with help from an automated research system. It beats its base model on most research and agent benchmarks while activating only about 3B parameters per token.
Baidu's 3B open-source vision-language model parses hundred-page documents in a single request using a sparse attention trick that keeps memory flat.
Qdrant's ONNX port of multilingual-e5-large hits 926K downloads, giving teams a lightweight way to run a top multilingual embedder in production.
CopilotKit released an MIT-licensed template for proactive AI agents that handle work in the background, request approvals, and run across web, mobile, Slack and Teams.
SemLoc turns LLM program reasoning into runtime-checkable constraints, boosting Python fault localization Top-1 accuracy from 6.4% to 42.8% over spectrum baselines.
AtomicChat strips refusals from DeepSeek V4.1 Flash and ships 1.75 to 2.25-bit quants that run a 763B MoE on a single 12 GB GPU.
A Carnegie Mellon and Browser Use benchmark of 100 hour-long live-web tasks exposes how far web agents still are from reliable knowledge work.
Artificial Analysis rebuilds its flagship benchmark with 60% private tests, two new agentic evals, and a widening open-weights gap.
A new objective-based Self-Organizing Map variant gives every data point its own continuous latent coordinate, cutting cost from O(NM²) to O(NM) while topping 16 benchmarks.
A 27B Qwen fine-tune distilled on 200K coding and agent trajectories, shipped as GGUF for local inference with llama.cpp, Ollama, and vLLM.
NVIDIA released Auto Ontology, an open-source Python framework that builds semantic graph layers over enterprise data so agents generate SQL deterministically with provenance.
A Berkeley team trains a residual RL policy to nudge human teleoperators off-course during demos, producing datasets that fix the compounding-error problem in robot learning.