Superpower Daily
Artificial Analysis Rebuilds Image Editing Arena With Different...LlamaIndex and Kaggle Launch a Document Benchmark for Cost and...Lyte Raises $165M at $1.6B Valuation for Robot PerceptionAbliteration Removes Refusal Mechanisms for Offensive Cyber WorkNew Mexico Lets Districts Seek Approval to Leave K-2 AI Reading...HPE Raises 2026 Revenue Growth Forecast to 34%-37% After $12.21...Artificial Analysis Rebuilds Image Editing Arena With Different...LlamaIndex and Kaggle Launch a Document Benchmark for Cost and...Lyte Raises $165M at $1.6B Valuation for Robot PerceptionAbliteration Removes Refusal Mechanisms for Offensive Cyber WorkNew Mexico Lets Districts Seek Approval to Leave K-2 AI Reading...HPE Raises 2026 Revenue Growth Forecast to 34%-37% After $12.21...

Full archive

Open release: Latest AI News

Search reporting on AI markets, models, products, companies, policy, and culture.

Category

More filters
Open release
Showing 19-36 of 45

Last week

15 stories
Open releaseTools

Intel Releases 20 Agent Skills for Arc GPUs, With an Open-Port Security Catch

The open repository turns Intel’s GPU deployment steps into reusable instructions for coding agents, but several model-server launchers bind broadly without authentication by default. Intel has released 20 open-source Agent Skills that give coding agents procedures for getting AI workloads running on Arc and Arc Pro GPUs. XPU is Intel’s GPU software path in this collection.

3 min readRead story
Open releaseModels

Tencent Releases Open-Source Hy4, a 770B Model That Activates 49B Parameters Per Request

Tencent is pairing a large mixture-of-experts model with plans for product integration, while warning that the preview can spend too long working through difficult questions. Tencent has released Hy4 preview, an open-source model aimed at software engineering, research, and financial analysis. One model, two parameter counts Hy4 uses a mixture-of-experts architecture with 770 billion parameters in total.

2 min readRead story
Open releaseTools

Visa’s VVAH Adds AI Patch Writing Across 11 Stages, Leaving Merge Decisions to Humans

The open-source harness now moves from detection through candidate remediation and adversarial validation. Its default can alter a repository’s working copy, while build, test, review, and merge controls remain outside the tool. The consequential boundary is that VVAH edits source files by default, but Visa says it does not merge changes or replace an organization’s normal build, test, and review flow.

3 min readRead story
Open releaseInfrastructure

Snowflake Open-Sources Semi-Persistence With Sub-Second Single-GPU vLLM Swaps

The release treats GPU model copies as disposable while retaining weights in CPU memory, aiming to let many specialized models share costly accelerator capacity without long reload delays. Idle models can give back GPU memory only if they can return before the next request waits too long. Avoiding the slow path on every swap The target is a tradeoff in vLLM, the inference engine Semi-Persistence currently wraps.

3 min readRead story
Open releaseTools

Experiential Open-Sources an Agent Router That Learns From Traffic, but Results Are Unverified

The project turns an agent’s past requests into evidence for model selection and training, but its cost and quality gains remain company-run benchmark results. Experiential offers a familiar promise: one endpoint for many model providers. It also makes a harder claim—that its gateway can learn from an agent’s production traffic and steer future work toward a better balance of cost and quality.

3 min readRead story
Open releaseModels

Z.ai Releases 320B GLM-5.3-Flash With MIT Weights, 1M Context and Low API Rates

The model gives developers a permissively licensed route to long-context, image and video workloads, while its performance and serving-efficiency claims remain company-reported results. Z.ai has released GLM-5.3-Flash, a 320-billion-parameter mixture-of-experts model that activates 18 billion parameters for each token. The August 26 release is the first natively multimodal model in Z.ai’s GLM-5 series.

2 min readRead story
Open releaseModels

IBM Releases Granite 4.2 Open Weights, Reserving Agentic Training for 8B and 30B

The new family gives self-hosted deployments a choice between a smaller tool-capable model and larger variants trained for code, search and other tool-use tasks. IBM has released Granite 4.2, an open-weight language-model family built for download and self-hosting. The central distinction is not just size: its 8B and 30B models receive specialized training for tool use, while the 3B version does not.

2 min readRead story
Open releaseInfrastructure

AWS Publishes a Phone-Ordering AI Pattern That Connects Restaurant Data to Claude

The sample shows how a voice agent can take an order from a phone call through backend tools, but restaurants must supply and operate the systems holding their own menus, carts, orders and locations. AWS has published a deployable reference architecture for a restaurant AI host that answers phone calls, handles menu questions, recommends pickup locations, manages a cart and confirms an order aloud.

3 min readRead story
Open releaseTools

Liquid AI’s Pipette Tests 1,000+ On-Device AI Setups—and Limits Cross-Device Rankings

The open-source suite puts model, quantization, runtime and hardware in one result. Its quality scores still come from H100 reference systems, while early results are not designed for cross-device comparisons. Liquid AI has released Pipette, an open-source benchmarking platform built around a consequential premise: on-device AI behavior belongs to the full deployment configuration, not the model alone.

3 min readRead story
Open releaseTools

Nvidia’s NeMo Switchyard Routes Agent Calls Across Models, Not One Default

The open-source routing layer can switch models as an agent encounters tool results and errors, but companies must now test the routing policy itself as prices and capabilities move. Nvidia has released NeMo Switchyard, an open-source Rust-based proxy and library that routes large-language-model traffic across models and providers. A router is not the agent manager The boundary matters.

4 min readRead story
Open releaseModels

MiniMax Maps H3 to 24GB PCs and Server Stacks, but Keeps Context Layer Hosted

The index gives developers more routes to run and adapt H3, yet the layer that turns mixed media into model-ready context remains an API product—and local use is restricted in several major markets. MiniMax has published an H3 integrations index that points developers from a 24GB local setup to ComfyUI workflows, Apple Silicon, fine-tuning, and multi-GPU serving through SGLang and vLLM-Omni.

3 min readRead story
Open releaseInfrastructure

AWS Backs ARD Open Spec to Federate AI-Agent Discovery Across Environments

The Apache 2.0 specification is designed to make agents, MCP servers, tools and APIs discoverable across separate catalogs without moving their governance into one central system. As companies spread AI agents and their tools across clouds, SaaS products and internal systems, finding an approved resource can become a manual integration job. Agent Registry is currently in preview.

3 min readRead story

August 2026

3 stories
Open releaseTools

Georgia Tech’s IPO-Mine Organizes IPO Filings; Models Can Disagree With Experts on Charts

The accompanying preprint finds IPO prose is becoming more uniform while charts and infographics grow more varied, making visual review a rising constraint on automated disclosure analysis. Georgia Tech’s Financial Services Innovation Lab has developed IPO-Mine, an open-source toolkit and dataset for long, multimodal IPO documents. IPO filings are disclosures submitted to the U.S.

3 min readRead story
Open releaseModels

MiniMax H3 Opens Its Video Weights, but Not Where Many Developers Work

The release gives smaller eligible organizations a route to self-hosted video generation, but the highest-resolution system and global access remain a hosted-service proposition. MiniMax H3 promises a unified route from prompts and reference media to video with synchronized stereo sound. One model, mixed inputs Released July 31, H3 accepts text, images, video and audio as a single input context.

3 min readRead story
9 more next