Tech News
← Home  ·  All topics

Apache 2.0

9 GoKawiil briefs on this topic

New AI lab releases Inkling 2 Small, a 594B-parameter mixture-of-experts model

The developer unveiled Inkling 2 Small, a Mixture-of-Experts model with 594 billion total parameters and 17 billion active, supporting image inputs and variable effort levels. The model scored 31 on the Artificial Analysis Intelligence Index, a notable jump from the company's prior generation, with gains attributed to more efficient pre-training and a unified reinforcement learning run. It is currently accessible only through the Tinker platform with closed weights, ahead of a planned Apache 2.0 open-weight release next week pending safety testing.

BetterWispr launches as free open-source dictation app for Mac

BetterWispr is a new Apache 2.0-licensed dictation tool for macOS 14 and later that requires no subscription, account or API key. It runs speech-to-text locally using downloadable models or Apple's on-device speech recognition, letting users dictate into any text field via a hold-to-record shortcut like Option+Space.

Google releases EmbeddingGemma 2, a multimodal on-device embedding model

Google introduced EmbeddingGemma 2, a 740M-parameter embedding model that maps text, code, images, video, and audio into a single unified embedding space. It supports flexible output dimensions from 768 to 128 via Matryoshka Representation Learning, offers an 8K context window—four times larger than the text-only EmbeddingGemma—and is released under the Apache 2.0 license.

Google releases EmbeddingGemma 2, adding multimodal embeddings on-device

Google launched EmbeddingGemma 2, a follow-up to its original EmbeddingGemma model that reportedly surpassed 20 million downloads. The new 740-million-parameter model is built on the Gemma 4 architecture, released under an Apache 2.0 license, and extends beyond text to embed code, images, video, and audio in a single shared space for on-device inference.

Nvidia joins Microsoft, Meta, IBM in open-weight AI letter

On July 24, Nvidia signed 'Open Weights and American AI Leadership' alongside Microsoft, Meta, IBM, Hugging Face and Mistral. The letter argues that U.S. AI leadership requires not just powerful frontier models but an open ecosystem that spreads AI capabilities across the economy, distinguishing open-weight models—whose parameters can be downloaded, run and modified under licences like Apache 2.0—from closed models accessed only via API.

AWS releases open-source Strands harness for building general-purpose AI agents

Strands harness is a preassembled agent harness that developers can run locally or deploy to cloud providers with a single line of Python or TypeScript code, released under an Apache 2.0 license. It is designed as a general-purpose agent rather than a coding-specific one, and its makers report it costs 28% less than comparable harnesses across six benchmarks while maintaining similar or better accuracy, and 77% less than Claude Code on the Fable 5 benchmark while scoring higher on Terminal Bench 2.1.

Bonsai 2 27B compresses a 27B model to 5.9GB with 98.2% performance retained

A new ternary-quantized model, Ternary Bonsai 2 27B, has been released, built on Qwen3.8 27B and using {-1,0,+1} weights with FP16 group scaling to shrink the model to about 1.76 effective bits per weight and a 5.9GB footprint. Despite being over 9x smaller than its full-precision counterpart, it retains 98.2% of aggregate benchmark performance across reasoning, coding, vision and agentic tasks, and supports a 262K-token context window under an Apache 2.0 license.

IFM releases K2 Horizon, a six-model open-source AI fleet with full training lifecycle

IFM has launched K2 Horizon, a family of six models ranging from 0.9B to 375B-A23B parameters, released under Apache 2.0 with datasets under licenses like ODC-BY. Unlike typical open-weight releases, IFM is publishing the entire training lifecycle—intermediate checkpoints, data recipes, architecture details, training code, logs, and evaluations—alongside final weights. The smallest models (0.9B, 3.7B, 7B) reportedly set new state-of-the-art benchmarks for their size class.

Jujutsu creator Martin von Zweigbergk becomes CTO of East River Source Control

Martin von Zweigbergk, who built the Jujutsu version control system while at Google, has been named chief technology officer of East River Source Control, a startup developing next-generation version control platforms. He will oversee engineering there while continuing as a core maintainer of Jujutsu, which remains open source under Apache 2.0. ERSC's storage product is set to enter private beta later this month.