- Anthropic51%
- Google39%
- OpenAI5%
- Alibaba1%
- xAI1%
- Meta1%
The story so far
written Sep 30What is happening
The market for the top AI model by the end of 2026 determines the distribution of venture capital and enterprise contracts among the five leading frontier labs. Anthropic currently holds a 51% probability of winning the market, followed by Google at 39% and OpenAI at 6% [13]. OpenAI withheld its GPT-6 Astra model during the September 29 DevDay, citing safety concerns and a decision to slow the pace of frontier development [1].
What to watch
The release of GPT-6 Astra would settle whether OpenAI can reclaim the lead from Anthropic, which recently launched Claude Opus 5.5 and Claude Sonnet 5.5 on September 22 and 28 [4][5]. A public benchmark update from the LMSYS Chatbot Arena would provide the first independent comparison of these models against xAI’s Grok 4.7, which the company introduced on September 21 [2].
Read the full brief · how we got here
How we got here
OpenAI released GPT-6.1 Sol on September 29, describing the system as having intelligence levels for coding and professional tasks near those of the unreleased Astra model [3][7]. Anthropic claimed its September 22 release of Claude Opus 5.5 matches the performance of the Claude Fable 5.1 system it launched earlier in the month [4]. xAI asserts that its Grok 4.7 model is twice as fast as comparable systems [2].
References · 7
- [1]cnbc.com — OpenAI DevDay 2026 recap - CNBC
- [2]x.ai — News: Research, Product & Company Updates | SpaceXAI
- [3]openai.com — OpenAI Research | Release
- [4]anthropic.com — Newsroom \ Anthropic
- [5]support.claude.com — Release notes | Claude Help Center
- [7]worthview.com — OpenAI DevDay 2026: All the Biggest Announcements, From Dots and GPT-6. ...
- [13]polymarket.com — Top AI Model Predictions & Real-Time Odds | Polymarket
Timeline
newest firstWhen we started following: OpenAI holds back Astra model as Anthropic and xAI release new frontier systems in September
OpenAI withheld its latest GPT-6 Astra model during its September 29 DevDay, citing safety concerns and a decision to slow the pace of frontier development [1]. The company instead released GPT-6.1 Sol, which it describes as having "near-Astra intelligence" for coding and professional tasks [3]. Anthropic launched Claude Opus 5.5 and Claude Sonnet 5.5 on September 22 and 28, claiming Opus 5.5 matches the performance of its Claude Fable 5.1 system released earlier in the month [4][5]. xAI introduced Grok 4.7 on September 21, asserting it is twice as fast as comparable models [2]. These releases impact the competitive standing for the best AI model by the end of 2026, a market where Anthropic currently leads at 51% followed by Google at 39% and OpenAI at 6%.