The week of July 22, 2026, was defined by timelines, ambition, and ecosystem expansion. Elon Musk publicly committed to Grok 4.6 around August 7 and teased Grok 4.7 as a 2.1-trillion-parameter follow-up weeks later. Musk also pledged that Grok Imagine will produce a full-length film adaptation of Homer’s Odyssey by year-end. On the platform side, Grok Web shipped new enterprise connectors for Microsoft 365, Google Workspace, and X Ads, while Grok 4.5 quietly reached every SpaceXAI surface — grok.com, X, iOS, and Android — on the same day.

Here is our professional analysis of the developments that mattered this week.

Grok 4.6 Gets an August 7 Target — and Grok 4.7 Is Right Behind It

The centerpiece of this week’s xAI news was a single X post. On July 24, Musk wrote: “Grok 4.6 in 2 weeks and Grok 4.7 in 4 weeks.” By July 28, multiple outlets confirmed the clarification: Grok 4.6 is expected around August 7, 2026, with Grok 4.7 following a few weeks later.

What we know — and what we don’t

The parameter-count story is murkier than it should be. Musk’s July 28 post, as reported by American Bazaar, described Grok 4.6 as “the 1.5T model with significantly improved SFT & RL” — meaning stronger supervised fine-tuning and reinforcement learning on the same 1.5-trillion-parameter foundation as Grok 4.5. However, secondary commentary and a Facebook post by M.A. Ostrega on July 24 referenced Grok 4.6 as a “2T model” that is “better than our 1.5T in every way” — implying a 2-trillion-parameter upgrade. The AirEiter technical blog noted on July 27 that official SpaceXAI documentation does not yet list Grok 4.6 anywhere, and that some circulating figures may conflate Grok 4.6 with Grok 4.5’s 1.5T V9 foundation.

Grok 4.7 is more consistently reported: a 2.1-trillion-parameter model shipping weeks after 4.6, positioned directly against frontier models from OpenAI, Anthropic, and Moonshot AI.

The API reality check

As of July 27, Grok 4.5 remains the latest documented model in SpaceXAI’s official docs. There is:

  • No public model ID for Grok 4.6 on docs.x.ai
  • No pricing row in the API documentation
  • No context window specification
  • No benchmark scores
  • No statement on vision, audio, or open weights

The real signal that Grok 4.6 has shipped will be a grok-4.6 entry appearing in the model listing at docs.x.ai/developers/models and in the /v1/models API endpoint. Until then, Musk’s “two weeks” is a marketing commitment, not a product launch.

Why it matters

The accelerated cadence — Grok 4.5 on July 8, 4.6 targeting August 7, 4.7 in late August — signals SpaceXAI’s intent to ship roughly one major model per month through end of 2026. That pace puts xAI in direct competition with OpenAI’s GPT-5.x series and Moonshot AI’s Kimi K3, both of which are iterating on similar timelines. The strategic question for enterprise buyers is whether speed of releases translates to quality improvements on benchmarks that matter for production workloads — or whether xAI is optimizing for announcement frequency over capability leaps.

The SFT and RL improvements Musk highlighted are significant if they materialize. Stronger reinforcement learning from human feedback could close the gap on instruction-following and safety alignment, areas where Grok has historically lagged behind Anthropic’s Claude. But without published benchmarks or third-party evaluations, enterprise teams have no basis for evaluation until the model is actually available.

Grok Imagine Pledges Full-Length Odyssey Film by Year-End

On July 22, Musk posted on X: “Before this year ends, Grok Imagine will make a full-length movie of The Odyssey that is historically accurate and true to the art of Homer.” The announcement was made by quoting a roughly three-minute conversational scene between Odysseus and Calypso generated by Grok Imagine, shared by X account Heavy Pulp.

The capability gap

Grok Imagine Video 1.5 is the current generation, generally available in the xAI API as grok-imagine-video-1.5. According to documentation and technical analysis from Roborhythms, in July 2026:

  • Grok Imagine raised its maximum clip length from 10 seconds to 15 seconds, then quietly restricted access to the longer format for many accounts within hours
  • A 1080p toggle appeared but was effectively upscaled from 720p, not a true 1080p render
  • xAI shifted to a shared weekly quota covering Chat, Imagine, Voice, and Build — meaning even failed or rejected clips consume your allocation
  • The free plan offers no image or video generation at all

A full-length film is approximately 90–120 minutes. At 15-second clips, that would require 360–480 generated segments minimum, assuming zero retakes. The arithmetic alone makes the year-end pledge aggressive — and that is before accounting for narrative coherence, character consistency, audio synchronization, and editorial assembly across hundreds of clips.

The leading indicator to watch

The path from 15-second clips to a feature film runs through incremental clip-length increases: 15s → 60s → 5min → longer. If Grok Imagine pushes clip lengths to 60 seconds or beyond in the coming weeks, the Odyssey pledge becomes plausible. If clips stay at 15 seconds, it does not. Watch for:

  • Sustained availability of 15-second clips (the brief restriction suggests instability)
  • Any announcement of 30-second or 60-second clip support
  • Improvements to character consistency across clips
  • Audio generation quality, particularly dialogue and lip synchronization

Enterprise implications

AI-generated video is a fast-moving capability that enterprises are watching for marketing, training, and internal communications. Grok Imagine’s Video 1.5 adds Projects, multiple agents, and library search for organizing video content — features that suggest SpaceXAI is building toward production workflows, not just novelty demos. The shared weekly quota model, however, is a friction point for teams that need predictable throughput. Enterprise users on SuperGrok or API tiers should evaluate whether the quota allocation supports their production volume before building dependencies on Grok Imagine.

Grok Web Adds Microsoft 365, Google Workspace, and X Ads Connectors

On July 23, the Grok Web changelog revealed a significant batch of new connectors and capabilities:

New connectors

  • X Ads — view and manage X ad campaigns, groups, and ads; creatives can be generated via Grok Imagine
  • Microsoft Teams (Microsoft 365)
  • Outlook (Microsoft 365)
  • Outlook Calendar (Microsoft 365)
  • Google Drive (Google Workspace)

New API features

  • App-chat text-to-speech via text-to-speech and tts-voices endpoints for Grok Web and app chat
  • The ENABLE_GROK_COMPUTER feature flag was removed, suggesting the computer-use capability has either graduated to general availability or been deprecated

Why it matters

These connectors transform Grok from a chatbot into an enterprise productivity assistant that can read your email, manage your calendar, search your drive, and run your X ad campaigns — all within a single interface. The Microsoft 365 and Google Workspace integrations are particularly significant because they position Grok against Microsoft Copilot and Google Gemini in the exact workflows where those products have their strongest distribution advantages.

For enterprise teams, the key question is data handling. Grok Web connectors require access to your Microsoft 365 and Google Workspace data — and SpaceXAI’s track record on data handling, particularly the Grok Build CLI scandal from two weeks ago, means organizations should evaluate these connectors with caution. The connector permissions model, data retention policies, and whether connector-sourced data is used for model training all need scrutiny before enterprise deployment.

The X Ads connector is strategically interesting for marketing teams. The ability to generate ad creatives via Grok Imagine and manage campaigns through Grok’s interface could streamline social advertising workflows — but it also deepens the dependency on SpaceXAI’s ecosystem, which carries its own risks.

Grok 4.5 Reaches Every Platform Simultaneously

On July 22, Grok 4.5 was deployed across grok.com, X, iOS, and Android on the same day — a notable departure from the staggered rollouts that have characterized previous model releases. This same-day multi-platform deployment suggests SpaceXAI has matured its deployment pipeline, reducing the fragmentation that previously left mobile users waiting for newer model versions.

For users, this means the Grok experience is now consistent across surfaces. For enterprises evaluating Grok as a platform, it signals operational maturity — the ability to coordinate releases across web, social, and mobile simultaneously is table stakes for a serious AI platform, and SpaceXAI has now cleared that bar.

Grok User Growth Stalls at 117 Million

WION reported on July 28 that Grok’s monthly active users tripled from 35 million in December 2025 to 117 million by March 2026, but growth has since plateaued. To combat the stall, xAI has been pushing Grok into new distribution channels:

  • A free Google Workspace add-on, lowering the barrier to adoption for enterprise users
  • The newly completed Grok Imagine video generation feature
  • The new enterprise connectors announced this week

What the numbers tell us

A tripling to 117 million users is impressive growth, but the plateau suggests Grok has captured the low-hanging adoption within Musk’s ecosystem (X users, tech enthusiasts, existing SpaceXAI customers) and now needs to compete for users who are not already in the SpaceXAI orbit. The Google Workspace add-on is a particularly aggressive distribution play — it puts Grok directly inside the workflow of users who are already paying for Google’s productivity suite, with no additional cost to try it.

The competitive landscape is unforgiving. OpenAI’s ChatGPT has a substantial head start in enterprise adoption, Microsoft Copilot is embedded in the world’s most widely used productivity suite, and Google Gemini is the default in Google Workspace. Grok’s differentiation — real-time data from X, the personality modes, the SpaceX engineering data pipeline — needs to translate into measurable productivity gains if the platform is to break out of its current user base.

What to Watch

  • Grok 4.6 launch — August 7 is the target. Watch for the model ID appearing in xAI’s API docs, benchmark scores, pricing announcements, and third-party evaluations. The gap between Musk’s “two weeks” and actual API availability will tell you how much lead time to build into your planning.
  • Grok 4.7 specs — The 2.1T parameter claim, if confirmed, would make Grok 4.7 one of the largest frontier models deployed. Watch for whether xAI publishes benchmarks or whether the parameter count remains the only headline number.
  • Grok Imagine clip length — Whether 15-second clips become consistently available (they were briefly restricted) and whether longer formats are announced. The Odyssey pledge’s credibility hinges on this trajectory.
  • Connector data policies — SpaceXAI needs to publish clear documentation on how data accessed through Microsoft 365, Google Workspace, and X Ads connectors is stored, retained, and whether it is used for model training. The Grok Build scandal makes this non-negotiable.
  • User growth trajectory — Whether the Google Workspace add-on and new connectors move the needle on the 117 million user plateau. The next reporting cycle will reveal whether the platform expansion strategy is working.
  • Grok 5 timeline — Still training on Colossus 2 in 6T and 10T parameter variants. Any update from SpaceXAI on Grok 5 timing would be significant, as it would indicate whether the monthly cadence extends into Q4 2026.

Follow Kevin Kaminski on X for daily AI ecosystem updates. Check back next week for the next xAI Weekly.