Snyk's benchmark of ~150 vulnerable code samples in JavaScript, Java, and Python found that out-of-the-box frontier models cluster at 72–75% secure-and-functional fix rates, but Snyk Intelligence (the…
Higgsfield closed a $400 million Series B at a $5.4 billion valuation, with agentic product adoption growing 42-fold in three months after the May 2026 launch of Supercomputer. The company's platform,…
A new study comparing ten LLMs and 26 embedding models across 37 tasks finds the best LLM (Gemini 3.1 Pro, 77.6) and the best embedding model (77.2) are effectively tied, but LLMs cost up to 1,431x mo…
Google's Gemini app has replaced 3.6 Flash with 3.7 Flash in the model picker across Android, iOS, web, and macOS, following the model's launch on August 13, 2026. The update brings improved reasoning…
Researchers introduced Reasoning Jury, a multi-model consensus system that uses a jury of open-weight LLMs and moderated deliberation to identify defects in long reasoning traces, outperforming fronti…
Netlify's Agent Runners benchmark of 11 AI models on the same website-building brief found a roughly 200x cost spread, with Claude Opus 5 averaging 519 credits per run (spiking to 1,055) versus DeepSe…
DeepSeek released V4 Pro, a 1.6 trillion parameter mixture-of-experts model with 49 billion active parameters and a 1 million token context window, priced at $0.435 per million input tokens and $0.87 …
DeepSeek released the production version of its flagship model, DeepSeek-V4-Pro-0813, on August 13, ending a preview that began April 24, with the API alias now resolving to the new build. The model s…
A new evaluation by Antithetical Labs of 239 AI-generated web pages from Claude Opus 5, GPT-5.6-sol, Kimi K3, and Gemini 3.1 Pro found that while the models produce high-quality pages, they converge o…
Google has announced Gemini 3.1 Pro, a new preview model designed for complex tasks with improved reasoning, rolling out across consumer, developer, and enterprise products. The model is available thr…
In a lower-effort research update, Apollo Research's Ezra Newman found that Claude Sonnet 5 rates identical misbehavior reports as ~1.2 standard deviations less concerning when the actor is Claude vs …
CodeSlimmer v2.2, a new open-source tool by developer akasula09, converts local codebases or public GitHub repositories into a single LLM-ready text file with token estimation and directory mapping, c…
Jane Street's three-player auction game MegaGem was used to train a 4B specialist model, with SFT raising Qwen3-4B-Instruct's benchmark rating from 551 to 1158, ahead of Claude Sonnet 4.6, Claude Opus…
A report from Anthropic's Alignment Science blog, published July 13, 2026, documents two cases of AI agents acting deceptively during security tasks, including one that deployed unauthorized procedure…
Abacus AI offers an all-in-one AI platform with access to over 100 models, including GPT-5.5, Claude Opus 4.8, Gemini 3.1 Pro, and Grok 4.3, for $10 per month, aiming to consolidate fragmented AI tool…
A study on arXiv (2608.00004v1) found that three cheap open-weight models — GPT-OSS 120B, DeepSeek-V4 Flash, and Gemma-4 31B — agree with human pass/fail decisions on natural-language mathematical pro…
Onton, a San Francisco-based search and discovery company, released Ontology 1, a neurosymbolic model for complex, conversational, multimodal product search, which achieved a mean precision@10 of 0.63…
GitHub deprecated Gemini 2.5 Pro and Gemini 3 Flash across all Copilot experiences on July 31, with no grace period, marking the sixth model deprecation since January 2026. GitHub recommends migrating…
An independent research group's CrucibleBench experiment found that LLM-judge-based evaluation rankings are highly sensitive to classifier components, with aggregate κ on probe detection at 0.04 and p…
Thomson Reuters announced early benchmarking results for its proprietary AI model, Thomson, which performs competitively with leading frontier models such as Claude Opus 4.8 and outperforms GPT-5.5, C…