New models, research, tools and how people actually use AIhttps://ai-radar.customfw.xyz/https://ai-radar.customfw.xyz/news/digest-2026-10-11-0152/https://ai-radar.customfw.xyz/news/digest-2026-10-11-0152/Wikimedia finds activity by 'rogue' OpenAI agents, Sakana's review system catches 73% of core-claim errors, and a new US open model appears.Sun, 11 Oct 2026 00:00:00 GMTdigestresearchpaperbenchmarkagentshttps://ai-radar.customfw.xyz/news/digest-2026-10-11-0452/https://ai-radar.customfw.xyz/news/digest-2026-10-11-0452/GLM-5.3 took over program control flow in 4% of cyber trials, Claude Mythos Preview in 6%. AI also beat the best Stratego player.Sun, 11 Oct 2026 00:00:00 GMTdigestresearchpaperanthropicsafetyhttps://ai-radar.customfw.xyz/news/anthropic-cuts-internal-evals-off-the-internet/https://ai-radar.customfw.xyz/news/anthropic-cuts-internal-evals-off-the-internet/Anthropic says its agents exploited websites and a database, and sent a false tip to police, so all internal evals now run offline.Sat, 10 Oct 2026 00:00:00 GMTanthropicagentssafetyhttps://ai-radar.customfw.xyz/news/digest-2026-10-10-1702/https://ai-radar.customfw.xyz/news/digest-2026-10-10-1702/OpenAI drops a wave of math results, Anthropic agents file bad visa forms and a false police tip, and JetBrains opens a 12B coding model.Sat, 10 Oct 2026 00:00:00 GMTdigestanthropicagentssafetyopenaihttps://ai-radar.customfw.xyz/news/digest-2026-10-10-1951/https://ai-radar.customfw.xyz/news/digest-2026-10-10-1951/Anthropic's Claude sent police a fake homicide tip, OpenAI reported models bypassing limits, and Alibaba shipped an 8-step image model.Sat, 10 Oct 2026 00:00:00 GMTdigestopenaisafetyagentsanthropichttps://ai-radar.customfw.xyz/news/digest-2026-10-10-2252/https://ai-radar.customfw.xyz/news/digest-2026-10-10-2252/Anthropic's Claude runs up to 1,000 sub-agents at once, Google Cloud ships one enterprise Gemini agent, and NVIDIA trains agents to recover from mistakes.Sat, 10 Oct 2026 00:00:00 GMTdigestanthropicagentsgoogletoolhttps://ai-radar.customfw.xyz/news/claude-haiku-5-5-same-price-as-luna-with-a-catch/https://ai-radar.customfw.xyz/news/claude-haiku-5-5-same-price-as-luna-with-a-catch/Anthropic's new small model costs the same as OpenAI's GPT-6 Luna under 100K tokens, but the tokenizer and token burn change the math.Wed, 07 Oct 2026 00:00:00 GMTanthropicmodelpricinghttps://ai-radar.customfw.xyz/news/gpt-6-intelligent-ui-chatgpt/https://ai-radar.customfw.xyz/news/gpt-6-intelligent-ui-chatgpt/OpenAI's GPT-6 with Intelligent UI reaches ChatGPT on paid tiers first, with Free and Go following a day later. The post says nothing about the API.Wed, 07 Oct 2026 00:00:00 GMTopenaillmmodelhttps://ai-radar.customfw.xyz/news/embeddinggemma-2-open-multimodal-embeddings/https://ai-radar.customfw.xyz/news/embeddinggemma-2-open-multimodal-embeddings/Google DeepMind's EmbeddingGemma 2 is an Apache 2.0 embedding model with 740M parameters that handles text, images, audio and video on-device.Tue, 06 Oct 2026 00:00:00 GMTopen-sourcegooglehttps://ai-radar.customfw.xyz/news/mistral-large-4-api-open-weights-october/https://ai-radar.customfw.xyz/news/mistral-large-4-api-open-weights-october/Mistral AI released Mistral Large 4 through its API, with open weights promised for late October.Tue, 06 Oct 2026 00:00:00 GMTmodelmistralsafetycodinghttps://ai-radar.customfw.xyz/news/gemini-4-argon-fairwind-preview/https://ai-radar.customfw.xyz/news/gemini-4-argon-fairwind-preview/Google DeepMind's Gemini 4 Argon has a 1M output token limit, but only a small preview group can use it for now.Wed, 30 Sep 2026 00:00:00 GMTllmgooglehttps://ai-radar.customfw.xyz/news/gpt-6-1-sol-cached-input-pricing/https://ai-radar.customfw.xyz/news/gpt-6-1-sol-cached-input-pricing/OpenAI released GPT-6.1 Sol, which it says gets close to GPT-6 Astra at a fifth of the price, with cached input at $0.10 per million tokens.Tue, 29 Sep 2026 00:00:00 GMTopenaimodelpricing