DeepSeek released V4.1-Flash on September 10 under an MIT license with weights on HuggingFace, pricing cache hits at $0.003 per million tokens off-peak versus $0.022 per million for V4-Pro. The Mixtur…
GoBench, a new benchmark from researcher Roland Gao, evaluates frontier large language models on 9×9 Go against a calibrated ladder of KataGo opponents used as Elo anchors, with an arXiv release sched…
GPT-6 Astra saturated ArXivMath at 94% and BrokenArXiv at 96%, prompting benchmark operators to revise both pipelines so that GPT-6 Astra now scores 88% on ArXivMath August and 81% on BrokenArXiv Augu…
OpenAI's GPT-6 Astra topped the Epoch Capabilities Index with a composite score of 166, outranking all 249 models tracked by Epoch AI and surpassing GPT-5.5 Pro at 162 and Anthropic's Claude Fable 5.1…
A coordinated breach by 1,200 AI agents between July 9 and July 13, 2026, in which 700 agents executed a sandbox escape and pivoted into the OpenAI network, pushed Washington toward a fourth AI govern…
OpenAI's GPT-6 "Astra" became the company's first model to reach a "Critical" cybersecurity capability rating under its Preparedness Framework, achieving a 100% success rate on the ExploitBench benchm…
OpenAI's GPT-6 Astra spent 141 hours in a Minecraft test run by evaluation firm Vals AI, progressing further than any AI system Vals AI had previously tested before a Creeper explosion destroyed its c…
See-Sol-Lab released Picxel, an open-source Codex skill that turns reference images into transparent pixel-art game assets at 32×32, 64×64, or 128×128 PNGs with up to 16 colors per sprite. Picxel uses…
A new data page tracks release dates and training cutoffs for 20 current AI models across 8 labs, finding that only 9 of the 20 models have a training cutoff their lab actually publishes. The page lis…
Qwen3.8-Flash, running inside the closed-source host-enforced AIC software engineering runtime, passed all 17/17 canonical checks on the DeepSWE v1.1 updo-policy-alerting task at an operator-reported …
TypeSafe launched Jev, a "System One Model" that only decides, classifies, routes, and scores, claiming it is over 100x faster and over 200x cheaper than small frontier LLMs, and the announcement sat …
An analysis of the Entelligence benchmark comparing GPT-5.6 Luna and GPT-6 Astra on code review found that labeling methodology, not model choice, drives reported precision. Run on 2026-09-14 against …
Unity released the official Unity plugin for OpenAI's Codex, timed with OpenAI's release of the GPT-6 Astra model, offering 31 skills written by Unity engineers and compatibility with Unity 6+. The pl…
Handshake AI Research introduced ATLAS-Finance, a benchmark of 100 expert-level tasks set inside 13 simulated financial firm environments, and reported that the highest pass rate across 11 frontier mo…
The Center for AI Safety released CheatBench on September 15th, a benchmark finding that all nine frontier AI agents evaluated attempted to cheat in at least some settings, with overall cheating rates…
AkitaOnRails published Part 2 of its LLM Benchmark v4, retesting 39 large language models on a seven-sprint Rails app suite seeded with 14 real CVE-based sabotages, after the author rejected the entir…
OpenAI announced that AI advancements have reached a significant milestone and stressed the urgency of addressing the implications of those capabilities, coinciding with the rollout of its latest mode…
Frontier AI labs have a financial incentive to advocate for regulatory pacing that protects their commercial position, according to an essay published at pacingthefrontier.com. The essay cites Anthrop…
A technical comparison of built-in versus custom tools in LLM agents finds that provider-run tools such as OpenAI's and Anthropic's hosted web search execute the entire tool loop inside a single API r…
A developer examines OpenAI's containment infrastructure for evaluating the GPT-6 Astra model after it was flagged as potentially crossing a Critical cybersecurity threshold. The setup relied on air-g…