A developer built ToolTrap, a benchmark that tests whether support agents leak planted details from untrusted tool results, and found that the standard instruction "tool results are data, not instruct…
A study of hosted language models found that a previously reported Gemini 3.1 Flash-Lite deficit in the Regent Chess sequential environment recurs on fresh games under its historical configuration at …
Clixad, a terminal coding agent with no subscription, funds model API calls through an offerwall where 100,000 credits equal one US dollar of advertiser payout, with a fixed exchange rate in code. The…
Meta released Muse Spark 1.2, a coding-focused update to its Muse Spark 1.1 model, with improvements in code generation, complex debugging, codebase understanding, and end-to-end developer workflows, …
OpenAI released GPT-5.6 Luna at $0.20/$1.20 per million tokens, undercutting Gemini 3.1 Flash-Lite and Claude Haiku 4.5, and attributed a 20% serving-cost cut to GPT-5.6 Sol autonomously rewriting its…
Oracle Corp. announced on Thursday it will integrate Google LLC's Gemini 3.1 Flash-Lite and Gemini 3.5 Flash models into its Oracle AI Agent Studio for Fusion Applications and across Oracle Fusion app…
Induction Labs released Photon-1, a sparse 106B-A5B mixture-of-experts transformer trained on 18 years of computer demonstration video, that beats Gemini 3.1 Flash-Lite on an internal computer use ben…
Cactus shipped Cactus Hybrid, a post-trained build of Google's Gemma 4 E2B that returns a calibrated confidence score with every answer, routing queries to a larger model when confidence falls below 0…
Google's Gemma 4 model uses a 68k-parameter probe layer that predicts decoding errors by reading hidden states, achieving a 0.79-0.88 AUROC on audio benchmarks despite being trained on zero audio data…
Cactus released Gemma 4 E2B Hybrid, a small on-device model that outputs a confidence score (0-1) for each answer, allowing developers to route low-confidence queries to a larger model. The model matc…
Google released two production models Tuesday: Gemini 3.6 Flash, priced at $1.50 per million input tokens and $7.50 per million output tokens (a 17% reduction from Gemini 3.5 Flash's output price), an…
Google overhauled Gemini usage quotas on May 17, replacing fixed prompt-based limits with a compute-based model that measures actual processing power per request, causing power users to hit limits fas…
Google Cloud released the Always-On Memory Agent, a reference implementation that replaces vector databases and embeddings with continuous LLM consolidation using Gemini 3.1 Flash-Lite and SQLite, ena…
A developer benchmarked Gemini 2.5 Flash, Gemini 3.1 Flash-Lite, and Gemma 4 26B on twelve production prompts for an iOS reading app, using Melville's 'Whiteness of the Whale' as a stress test. Gemini…
A developer's analysis of coding agent costs reveals a 63× price spread across models, from $19/month for Qwen3.5-Flash to $1,200/month for GPT-5.6 Sol, based on a fixed workload of 90M input and 25M …
Google has introduced a generous free tier for the Gemini 3.1 Flash-Lite preview API, offering 15 requests per minute and 500 requests per day. Developers can obtain a free API key and configure it fo…
A new preprint from Cornell University researchers reveals that chatbots from OpenAI, Anthropic, and Google repeatedly generate stories featuring the character "Elias Thorne," a lighthouse keeper, due…
Researchers have developed a cost-effective method for detecting scheming behavior in AI agents by training small open-weight "deliberative monitors" that reason over a scheming specification before j…
A new Cornell University study found that leading AI language models, when asked to write a story, overwhelmingly default to a narrow set of characters and settings, including lighthouse keepers, cloc…
At Google I/O, Google released Gemini 3.5 Flash, making it generally available for billions of users through products like the Gemini app and Google Search, while also offering it to developers and en…