🤖AI Newsletter
September 25, 2026 · 10:31 Uhr
1Open-Source AI: AT&T, Airbnb & Deloitte switch to cheaper models
@Polymarket / New York Times Corporate America is undergoing a strategic reversal: large corporations like AT&T, Airbnb, and Deloitte are massively migrating to open-source AI models to avoid the high licensing costs of OpenAI and Anthropic. The trend marks a fundamental power shift in the AI market away from proprietary vendors. OpenAI and Anthropic face significant pricing pressure, while open-source providers like Mistral and Meta benefit.
2Salesforce + NVIDIA: Domain-specific CRM model 'Koa' launched
@Benchmarkitai Salesforce and NVIDIA have jointly unveiled 'Koa', a specialized AI model for CRM applications directly integrated into Salesforce workflows. The partnership illustrates the trend toward domain-specific models instead of general-purpose LLMs in the enterprise segment. For the market, this signals consolidation: platform giants are building vertical AI stacks that displace independent vendors.
3GPT-6 Astra: OpenAI sparks AGI debate – Anthropic counters with Opus 5.5
@Dev__Vishwajeet / @postiglionr According to X discussions, GPT-6 Astra has for the first time triggered serious AGI debates – however, Anthropic's Claude Opus 5.5 beats it on nearly all benchmarks. This is a significant escalation from the previously reported launch of both models: Opus 5.5's benchmark victory shifts market leadership and explains the 99% probability on Polymarket. The intensity of competition between the two top labs reaches a new peak.
4Anthropic MHS: AI agents directly control microscopes and robotic arms
@Dr_Singularity Anthropic has introduced the Model Hardware Standard (MHS), which enables AI agents to autonomously operate physical laboratory equipment like microscopes and robotic arms, allowing them to conduct independent scientific experiments. The standard could become the industry norm for physical AI integration, similar to USB for peripherals. Pharmaceutical, biotech, and manufacturing sectors face direct disruption of their laboratory processes.
5US-China AI race: Trump and Xi in direct confrontation
BBC News / @maxvaldes BBC reports on a direct clash between Trump and Xi in the escalating AI race, while Alibaba is considered the best Chinese AI company with 92% Polymarket probability. The geopolitical dimension of AI competition reaches a new diplomatic escalation level. For companies, the risk of export controls, chip sanctions, and divided global AI ecosystems intensifies.
Situation Report
In September 2026, the AI market is at a turning point: Anthropic's Claude Opus 5.5 dominates benchmark rankings and outperforms OpenAI's GPT-6 Astra, escalating competition between the two leading labs to a new level of intensity. Simultaneously, massive migration of large corporations to open-source models structurally undermines both vendors' business models, while domain-specific model partnerships like Salesforce/NVIDIA demonstrate that real value creation lies in vertical applications. Anthropic's MHS standard for physical hardware control marks the transition from digital to physical AI autonomy and opens billion-dollar industry and science markets. Geopolitically, the Trump-Xi confrontation intensifies the risk of global AI bifurcation that could sustainably burden supply chains, chip availability, and international cooperation.
Tokens: 2,134(1,302 in · 832 out)
