xAI expands Grok 4.5 across web, X and mobile, teases Grok 4.6

xAI is scaling distribution of Grok 4.5, first launched July 8, across all its surfaces — grok.com, X, and dedicated iOS/Android apps — and leaning into a value-for-money pitch. Musk tweeted flatly that 'Grok 4.5 is the best value for money AI,' with the model positioned as 'Opus-class' in capability but priced aggressively at $2 on the API and touting high token efficiency.

The more interesting move is productization: new 'Grok Build' features automate complex office workflows, generating multi-sheet Excel files, presentation slides, PDF summaries, and shipping native Microsoft 365 add-ins — including an Excel plug-in that drew bemused reactions ('for some reason,' per TechRadar). This puts Grok squarely in the same lane as OpenAI, Anthropic and Microsoft, all of whom now ship Office/Excel add-ins, raising the question of whether role-specific tool integrations are a real differentiator or table stakes.

Musk also teased Grok 4.6, saying it will perform 'much better' on abstract-reasoning tasks like the ARC Prize, and separate reports point to a next-gen ~2-trillion-parameter Grok model nearing completion, benchmarked against Moonshot's Kimi K3. The rapid-fire 4.5 → 4.6 cadence fits the week's broader 'model-version-fatigue' theme, with several labs shipping point releases within weeks.

The skeptical take from developers: aggressive pricing and 'Opus-class' claims need independent benchmarks to hold up, and the flurry of Office add-ins reads more like feature-parity catch-up than a distinct advantage. What to watch: whether the promised ARC gains in 4.6 materialize, and whether the 2T-parameter model ships on a timeline that keeps pace with Alibaba's Qwen 3.8 Max and Chinese open-weight rivals.