xAI launches Grok-2 and Grok-2 mini in beta with improved reasoning, available to X Premium and Premium+ users, and plans to add them to its API later in August
Elon Musk-owned launched Grok-2 and Grok-2 mini in beta today with improved reasoning. The new Grok AI model can now generate images …
TechCrunchIvan Mehta
Context & Ripple Effects
This release follows X’s earlier plan to expand Grok beyond Premium+ to all Premium subscribers, using the social platform’s paid tiers as an initial distribution channel. Grok-2 adds a smaller companion model and image generation while keeping that subscriber-led rollout.
The planned API route matters because it opens a second path beyond X subscriptions. Later coverage shows that path materialized with an API for Grok 3 and Grok 3 Mini, making Grok’s progression from in-product feature to developer offering clearer.
First-order effects
X Premium and Premium+ users gain beta access to Grok-2 and Grok-2 mini, including the reported reasoning and image-generation improvements.
xAI prepares to make the models available to developers through an API, extending access beyond X’s paid audience.
Second-order effects
X’s subscription tiers become a testing and distribution layer for xAI’s models, tying perceived AI capability more closely to the value of Premium access.
An API launch would place xAI in direct competition for developer model workloads, where a smaller model can serve lower-cost or lower-latency use cases alongside the flagship model.
Third-order effects
The rollout points to a two-channel AI business model: consumer subscriptions provide embedded distribution, while APIs create a separate developer market and usage-based monetization path.
If repeated across model generations, rapid reasoning-model updates may make sustained access to compute, distribution and developer tooling—not a single launch—the central competitive advantage.
The trend: This is one step in the convergence of consumer-platform AI distribution and developer-facing model APIs, with reasoning capability becoming a recurring product tier rather than a one-off feature.
Woah, another exciting update from Chatbot Arena❤️🔥 The results for @xAI's sus-column-r (Grok 2 early version) are now public**! With over 12,000 community votes, sus-column-r has secured the #3 spot on the overall leaderboard, even matching GPT-4o! It excels in Coding (#2), [im…
much of grok-2 was coded in a conference room that barely fits 30 people. today @xAI shares its newborn baby with the world! incredibly proud of the work this team has done.
Here is (likely) the new AI model that Twitter is not (since a “settlement” last week) not using EU user data anymore to train - as they are already done training the model... Good job @DPCIreland! 🦾😳 #GDPR #AI
Never underestimate a small group of dedicated people to achieve something great. Proud to be a part of the hardest working team in AI. This is a huge milestone, and we're only just beginning.
Have been post-training Grok2 for a while and am excited to share that it's officially out!! We've been testing early versions of Grok2 on LMSYS chatbot arena under the names of sus-column-r and column-r, and the models perform well! @xai is still a super small team (an order
There are now five GPT-4 class models: GPT-4o, Claude 3.5, Gemini 1.5, Llama 3.1, and now Grok 2. All of the labs are saying there is room left for continued giant improvements, but we haven't seen any models truly leap above GPT-4... yet.