OpenAI retired the o3 model family from ChatGPT on August 26, 2026, ending a 90-day sunset period that began with the May 28 announcement, forcing developers and enterprises to migrate workflows to th…
JetBrains has released Junie Local, an on-device version of its AI coding agent that runs entirely on supported Macs without cloud inference or token charges. The tool uses a JetBrains-tuned, 4-bit qu…
JetBrains introduced Junie Local, an on-device coding agent that runs entirely on a Mac with no cloud usage or credits, shipping with a tuned Qwen3.6-27B model at 4-bit requiring an M5 Mac with 64 GB …
Researchers introduced OraRL, a reinforcement learning method for video multimodal large language models that uses annotations as oracle rollouts to improve sample efficiency and scalability. OraRL re…
AstraZeneca researchers developed an LLM-as-a-Judge evaluation framework for ChatInvent, an agentic drug discovery assistant, achieving human-aligned scoring with a majority-vote agreement of 0.86 aft…
OpenAI's latest model, GPT-5, achieves a 92% accuracy on the MMLU benchmark, surpassing human experts and setting a new state-of-the-art. The company claims this advancement will revolutionize AI appl…
Microsoft researchers introduced LoRA (Low-Rank Adaptation) in 2021, a parameter-efficient fine-tuning technique that reduces trainable parameters by approximately 10,000 times and GPU memory requirem…
Krasyn ran its Note Check tool against Omi Health's public scribe benchmark, testing 36 transcript-note pairs from six frontier models. The tool agreed with the benchmark's clean notes but disagreed o…
Aivolut Books, an AI-powered book-writing platform, is offering a lifetime Pro plan for $424.14 (regularly $2,376) with code SAVE15, enabling users to turn book ideas into complete manuscripts and pre…
Cursor Pro with Claude 3.5 Sonnet beat GPT-5 for daily coding tasks in a six-week comparison by a developer, winning 27 of 31 real tickets and averaging 23 minutes to a working pull request versus 41 …
In a blog post, an unnamed author argues that AI performance will not plateau, countering the common belief that progress is driven by brute-force scale with diminishing returns. The author contends t…
Researchers introduced DeltaML-Bench, a benchmark of 48 tasks from research papers that tests AI agents' ability to improve published baselines in real-world repositories. In evaluations, the ARG scaf…
A June 2026 paper by Lodha, Varnosfaderani, Chakraborty, and Mithal found that pruning an AI agent's context window to the last five tool calls raised task completion from 71.0% to 79.0% on a Microsof…
A new evaluation protocol called Fair-ASR, introduced in a paper, equalizes target-call budgets across jailbreak attacks and finds that rankings of 11 representative attacks shift dramatically dependi…
The developers behind the open-source OmnisBench project have released a benchmark designed to measure the true cost savings of LLM routing. Running 364 coding and math tasks across four models, they …
Researchers introduced ICD-Deepresearch, a DeepResearch workflow that combines structured EHR foundation models with language foundation models to forecast ICD codes for future clinical visits, achiev…
Researchers introduced Dripper, a lightweight framework for main HTML content extraction that uses small language models for constrained sequence labeling, achieving a throughput of 3.08 pages per sec…
A new benchmark called CIMemories, presented in a paper on arXiv, reveals that frontier large language models leak sensitive information from persistent memory in up to 69% of attribute-level violatio…
Roboflow's benchmark analysis shows OpenAI's GPT-5.6 Sol outperforms all previous OpenAI models on visual understanding tasks, including specialized vision models. The model handles diverse computer v…
Boards are now pressing startups on AI vendor concentration risk, asking founders to prove their business isn't dependent on a single model provider. OpenAI's August 2025 GPT-5 rollout briefly cut off…