OpenAI released ChatGPT Images 2.5 alongside GPT-5 Astra, a computer-use agent that can operate a desktop across applications, and in a demonstrated run a single prompt carried two source photos throu…
OpenAI released ChatGPT Images 2.5, an updated image generation and editing model available across desktop, mobile, and web on all tiers plus two API variants, adding sketch-to-image, improved identit…
OpenAI's ChatGPT Images 2.5, the successor to GPT Image 2, delivers its headline upgrade in frame-to-frame consistency, holding subject, style, and setting together across multi-step stop-motion edits…
AI agent harnesses — the tooling layer giving models file read/write and bash access — matter more to task completion than the underlying model, according to an analysis comparing Claude Code, Codex, …
DeepSeek released V4.1 Flash, a 552-billion-parameter mixture-of-experts model with 8 billion active input and 16 billion active output parameters that scored 74.2 on Deep SWE and 30 on Terminal Bench…
DeepSeek released DeepSeek V4.1 Flash, a 552 billion parameter mixture-of-experts model under an MIT license that activates only 8 billion parameters during prefill and 16 billion during decode, cutti…
Sustained AI inference workloads on Nvidia RTX 3090 and 4090 cards accelerate thermal paste pump-out, a failure mode in which paste migrates off the GPU die and causes clock throttling even while repo…
TokenRhythm released NeoHorse-1-9B, a 9-billion-parameter model post-trained from Qwen3.5-9B that scores a 69.04 macro average across ten benchmarks, a 3.44-point gain over the 65.60 Qwen3.5-9B baseli…
AI practitioners are advising developers to build model-agnostic agent skills — reusable instruction files, context documents, and plugins — so accumulated prompt and workflow expertise survives switc…
Three researchers — Beren Millidge, CTO of Zyphra; John Schulman, chief scientist at Thinking Machines and formerly of OpenAI; and Charlie O'Neill, head of model training at Baseten — argue that gener…
Anthropic's new documentation for Claude Opus 5 and Fable 5.1 advises users to hand over a complete task specification up front rather than a step-by-step script, warning that skills and prompts writt…
Higsfield's Genjutsu model applies a single AI-generated stylized still image onto recorded video to produce animated-style footage that preserves the original motion, expressions, and audio, accordin…
InclusionAI is offering free API access to its Ling-3.0-Flash-VL multimodal model and releasing the weights under the MIT license, the company said. The model uses a mixture-of-experts design with 124…
OpenBMB released MiniCPM5-2B, a 2.52 billion parameter open-weight language model that posts an average benchmark score of 53.9 across nine capability axes, ahead of Qwen3.5-4B (51.1), granite-4.2-3B …
RunPod Serverless lets developers deploy any Hugging Face model as an OpenAI-compatible API endpoint with per-second billing, scale-to-zero, and flash boot cold starts, according to RunPod. The platfo…
AI coding agents introduce security vulnerabilities both by writing insecure code directly and by installing third-party dependencies with known CVEs, often buried in sub-dependencies the agent never …
A blind benchmark of 50 one-shot website-generation prompts found Anthropic's Claude Fable 5.1 beat Fable 5 in human preference 30 times to 17, with 3 ties, and was favored by an AI visual judge 40 ti…
OpenAI released GPT Image 2.5, an updated image generation and editing model that ships in two variants, Flare (fast) and Sunburst (heavy), available to all ChatGPT, Codex, and API users. OpenAI's doc…
Hands-on testing of Google's Gemini video understanding model and the commercial detection API Sightengine found current AI video detection tools cannot reliably distinguish real footage from AI-gener…
Independent hands-on testing of DeepSeek V4.1 Flash, a preview agentic coding and reasoning model released by DeepSeek as an MIT-licensed FP8 checkpoint on Hugging Face under deepseek-ai/DeepSeek-V4.1…