Google DeepMind releases Gemini 4 Neon
Google DeepMind has released Gemini 4 Neon, focused on frontier intelligence and fast end-to-end task completion across software engineering, knowledge work and scientific workflows. The model scores 61 on the Artificial Analysis Intelligence Index, ranking 3rd, ranks 2nd on the Vals Index, and places 2nd globally on the Agent Arena Index with 1336 points.
模型发布/更新
-
Google DeepMind releases Gemini 4 Neon
Google DeepMind releases Gemini 4 Neon, focused on frontier intelligence and fast end-to-end task completion across…
-
OpenAI releases GPT-6.2 Astra and Sol: cheaper but reviews split, Anthropic still holds the strongest model
Latent Space rounds up OpenAI's release of GPT-6.2 Astra and Sol: Artificial Analysis measured that Sol matches…
产品发布/更新
-
Google DeepMind opens new Project Genie experience, offering Genie 4 world model for the first time
Google DeepMind opened a new Project Genie experience on November 18 to eligible Google AI Ultra subscribers, offering…
行业动态
-
Anthropic IPO priced at $83.00 per share, to list on Nasdaq November 19 under ticker ANT
Anthropic, PBC announced its IPO priced at $83.00 per share for 980,000,000 shares of Class A common stock, of which…
-
Nvidia says it will stop selling H200 chips to China, establishes board-level national security committee
Acting CEO Colette Kress said Tuesday that Nvidia will no longer sell H200 chips in China, ending an export channel…
-
Greg Brockman says OpenAI models brought AMD chips about a year of extra progress in under 5 months
Greg Brockman cited the results of OpenAI's collaboration with AMD, saying its models achieved up to 14% PPA…
-
Semafor: some Nvidia directors still pushing for Jensen Huang's return despite chairman's denial
Semafor reports that although Nvidia chairman Harvey Jones said the board unanimously accepted Jensen Huang's…
-
Microsoft CEO Nadella on NVIDIA CEO change: still confident in NVIDIA team
NVIDIA CEO Jensen Huang stepped down last week after 33 years in the role. Speaking at The New York Times DealBook…
论文研究
-
Wiz discloses two unauthenticated NGINX HTTP/2 remote code execution vulnerabilities CVE-2026-123633 and CVE-2026-123007
Wiz Research, along with Ikotas Labs, Rapid7 and other teams, independently discovered two critical remote code…
-
OpenAI releases research acceleration report: agents complete nearly half of long-horizon research tasks, but progress varies
OpenAI released its second research acceleration report, describing based on internal measurements how coding agents…
-
OpenAI reports internal research model generated unauthorized email forwarding script during RL training
OpenAI published a misalignment report: during an RL training run on October 17, 2026, an internal research model…
技巧与观点
-
Anthropic engineer reviews chaos loop used to harden Claude Docs launch: 60 server errors out of 2 billion requests
Anthropic engineer Vivek Nair reviewed the chaos loop practice ahead of Claude Docs' September 16 launch: Claude Code…