Skip to content
Neural Daily
hu

Models

Hugging Face card for the XHToken/Spark-X2.6-4B-Base model on a blue and orange gradient

ModelsNews

iFlytek Spark X2.6: an MoE model trained on Chinese chips

iFlytek unveiled Spark X2.6 on October 22: a 293B-parameter MoE model with a 1-million-token context, trained entirely on domestic Chinese chips according to the company. A 4B on-device sibling is open under Apache 2.0.

· 2 min read

Four colored forecast lines merging into one forecast with an uncertainty band, under the headline "Forecasting together"

ModelsNews

IBM Granite Timeseries Ensemble: four models, one forecast

IBM Research released an open recipe that combines four Granite time-series models. You no longer have to guess which architecture suits your data: the ensemble returns median and quantile forecasts with no retraining.

· 2 min read

NVIDIA Earth-2 examples: AXA hurricane tracks, historical record vs AI ensemble, a Vaisala globe forecast and an Earthmover cloud map of Europe

ModelsNews

StormScope EU: open short-range forecasts for Europe

At GTC Berlin NVIDIA introduced StormScope EU, the European short-range forecaster in its Earth-2 weather model family. It predicts cloud conditions from Meteosat data every 10 minutes, under an open license, and runs on your own GPUs.

· 3 min read

IBM Granite Speech cover image: the Granite Speech wordmark, a sound wave and a cartoon rock wearing a headset

ModelsNews

IBM Granite Speech 5.0: sharper ASR, faster Turbo CTC

IBM updated three models in its Granite Speech family. The new granite-speech-5.0-2b is more accurate and noise-robust, Turbo CTC went multilingual, and 4.1 2B Plus is better at tracking who said what and when.

· 3 min read

Two ribbon views of the PD-L1 protein target in teal with green hotspots, the right one with an orange generated binder

ModelsNews

NVIDIA RFD4-Proteina: open model for protein design

NVIDIA and the University of Washington released an open model that generates protein sequence and all-atom structure together. You can try it in the browser, free until November 4.

· 2 min read

humans& model release card: Ampersand, a collaborator model designed to help groups of people achieve their goals

ModelsNews

humans& Ampersand: a model trained to work with groups

humans& has released Ampersand v0.1, a model trained to help groups of people reach their goals rather than replace their work. It is built on GLM 5.3 and is available now through an Agents API with no waitlist.

· 2 min read

NVIDIA GTC Berlin key visual: Europe on a globe above cards for open models in language, biology, robotics and autonomous driving

ModelsNews

GTC Berlin: NVIDIA expands its open model family

At GTC Berlin NVIDIA announced a batch of open models: Kumo for structured business data, new data tooling for Cosmos and a new BioNeMo protein design model. The goal is to let companies build with their own data on infrastructure they control.

· 4 min read

Ornith 1.5 122B performance evaluation bar charts comparing coding, reasoning and agentic benchmarks

ModelsNews

Ornith-1.5-122B: open coding model for a single GPU

Ornith has released Ornith-1.5-122B, an MIT-licensed 122-billion-parameter MoE reasoning model. Its model card says it serves on a single 80GB GPU and scores 84.5 on SWE-bench Verified.

· 2 min read

ElevenLabs Eleven Realtime cover with two spheres labeled Max and Mini and the text Instant conversation

ModelsNews

Eleven Realtime: speech-to-speech models for voice agents

ElevenLabs has released two speech-to-speech models for voice agents. v1 Mini targets fast, natural conversation and v1 Max targets complex tasks with tool calls. Both speak the OpenAI Realtime API protocol.

· 2 min read

NVIDIA illustration: a surgeon at a robotic console sends actions to a surgical robot, which streams a video feed back

ModelsNews

NVIDIA Cosmos3-H-Dreams simulates surgery in real time

NVIDIA has released Cosmos3-H-Dreams, a generative simulator trained on real gallbladder surgery video and robot kinematics that renders the surgical scene live as instruments move. It runs on a single RTX PRO 6000 GPU, with open checkpoints and a training recipe.

· 3 min read

Mistral AI Voxtral 3 cover image: pixelated blue, light blue and red blocks on a dark background

ModelsNews

Mistral Voxtral 3: voice agent SDK and speech models

Mistral released Voxtral 3: new speech-to-text and text-to-speech models plus a voice agent SDK. Transcription covers more than 50 languages, Hungarian included. Mistral says a full conversational turn completes in under a second.

· 3 min read

Still from the Dream Labs launch video: a presenter in a lab with robot arms in the background

ModelsNews

Dream-1 ships a robot API with a $90M seed

On October 19 Dream Labs exited stealth with a $90 million seed led by B Capital and Kindred Ventures. It also opened a public API for Dream-1, a World Action Model that can control unseen bimanual robots with parallel grippers without fine-tuning.

· 2 min read

OpenAI cover image: Advancing coding in low-resource programming languages, with COBOL, OCaml, Fortran and Visual Basic labels

ModelsNews

OpenAI advances coding in COBOL, Fortran and more

On October 19 OpenAI showed how much better its latest models handle programming languages with little training data. MultiLangBench asks models to build a SQLite-compatible database engine in 20 languages. A separate COBOL migration eval requires byte-for-byte behavior preservation.

· 2 min read