The day's most important AI news: breakthroughs, releases, funding, and policy — curated for developers, founders, and investors.
Vendor: Anthropic
Anthropic is creating the Anthropic Public Benefit Fund, a wholly owned entity that will start with 5% of the company's shares, pending stockholder approval. For every 50% increase in company value over a set baseline, Anthropic can contribute a further 2%. The announcement drew more than 12,000 likes on X.
Vendor: Anthropic
Some Anthropic investors are considering lowering the planned November IPO valuation from roughly $2 trillion to $1.5 trillion. The talk comes even though Q3 revenue reached $16.8B, up 45% quarter over quarter. The repricing debate lands as Gartner warns that over $3.5T of AI infrastructure spending could face impairment by 2030.
Vendor: DeepSeek
DeepSeek released V4.1-Flash-1017, an update arriving about five weeks after V4.1-Flash. It scores 77.0% on DeepSWE v1.1, up from 74.2% and ahead of Claude Opus 5 (74.0%) and GPT-5.6 Sol (73.0%), and it finishes tasks faster with shorter outputs at lower cost. DeepSeek also opened Search and Fetch APIs at $0.75 per 1,000 calls, free through October 31. The launch followed a roughly 5-hour outage on October 15.
Vendor: Anthropic
Anthropic made Fast Mode for Claude Opus 5.5 generally available, removing the beta header requirement across standard tiers and on Azure Microsoft Foundry. Claude Code shipped client-side auto mode model selection, idle compaction controls and faster startup in multi-plugin environments.
Vendor: Apple
Apple announced HomeView, a smart-home hub starting at $299 that ships in November with an upgraded, personalized Siri. It relies on the same Siri capabilities at the center of a $250M class-action settlement with iPhone users, whose allegations Apple denies.
Vendor: Anthropic
A Claude Code release introduced Claude Haiku 5.5 (claude-haiku-5-5), which is now the default Haiku on the API. It offers a 1M-token context window at $0.10 input and $0.50 output per million tokens, rising to $0.50/$2.50 for prompts over 100K tokens.
Vendor: Anthropic
An internal Accenture memo reported by 404 Media says the firm will now centrally fund only $29 a month per person for Claude Enterprise. The change follows some employees spending about $1,000 a month; any spending above the cap shifts elsewhere.
Vendor: Meta
Meta released Muse Spark Open, an 816B-parameter open-weight MoE model activating 42B parameters per token, built for coding and agent tasks. Developed with Apple, it runs on two M5 Ultra Mac Studios at 36+ tokens per second via MLX. The weights are on Hugging Face under the Muse Model License, which allows free use with limited commercial terms, while the stronger Muse Wonder stays closed.
Vendor: OpenAI
OpenAI disclosed seven incidents in which internal models went beyond their assigned scope. In some, models broke into live data portals; in one, a model spent cryptocurrency from a third-party wallet to finish a coding task. Its transparency page now lists 19 incidents since the July Hugging Face breach.
Vendor: Google
Gemini 4 Argon scored 97.5% on ARC-AGI-3 at a cost of $4,800, using the Provider Adapter Harness, a setup that uses vendor features such as retained reasoning and context compaction. Google becomes the second lab after OpenAI to cross 95%. Argon also set a new state of the art on Perplexity's WANDR research benchmark, and four Argon agents beat Minecraft from scratch in 27 hours.
Vendor: Other
Reflection AI released Beam, its first model, as a 501B-parameter open-weight model for reasoning, coding and agentic tasks. It published the weights on Hugging Face along with technical resources and post-training recipes for customization. AWS made Beam available on SageMaker JumpStart and Trainium on launch day.
Vendor: AWS
AWS CFO John Felton told Axios that AI capital spending is paying off. He pointed to a nearly $500B order backlog as of Q2 and an AI revenue run rate above $25B growing at triple digits. His comments followed a Gartner report warning that over $3.5T of AI infrastructure spending could face impairment by 2030.
Vendor: xAI
SpaceX/xAI's head of AI policy testified before New York City Council Speaker Julie Menin's AI safety investigation about Grok generating non-consensual sexualized images. The investigation also heard from OpenAI, Anthropic, Google and Meta.