SpaceXAI has released its latest model, Grok 4.5 — the first since the company went public several weeks ago.

In a blog post published Wednesday, SpaceXAI characterized its new release as a workhorse that can tackle all of the typical tasks that the AI industry has sought to automate: coding and app-building, office and clerical work, research, writing, and other forms of routine knowledge work.

Grok can supposedly do all this for less spend, too, as SpaceXAI says that its model has “twice greater token efficiency” than other leading models. If it carries through to real-world use cases, that efficiency would be a big advantage for SpaceXAI, since the cost of tokens has been a growing concern for AI consumers.

The company released benchmark metrics Wednesday that appeared to show Grok’s competitiveness with other top models from SpaceXAI competitors, although just short of best-in-class:

In a post on his social media platform X (which is a subsidiary of SpaceXAI), founder Elon Musk compared the model to Opus, Anthropic LLM designed for intensive and complex tasks.

“Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow. It is an Opus-class model, but faster, more token-efficient and lower cost,” wrote Musk in a post on X.

Musk later added: “Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster. The combination of capability, faster speed and lower cost is what makes it competitive.”

SpaceXAI says that its new model costs $2 per million input tokens and $6 per million output tokens. That’s quite competitive, if Grok’s capabilities match SpaceXAI’s rhetoric.

Opus 4.7, by comparison, costs $5 per million input tokens, and $25 per million output tokens. OpenAI has tiered costs for different model versions: Sol, its most expensive, costs $5 for input tokens and $30 for output, while its least expensive, Luna, costs $1 for input and $6 for output.

It’s a big week for AI model releases. OpenAI is planning to release GPT 5.6, its latest, most powerful model, on Thursday. The release of that model had previously been limited by the Trump administration, due to concerns about its security implications. OpenAI has called it its “strongest model yet.”

Somewhere above the Great Plains, a virtual woodpecker is on its way to Alaska to deliver a message to my anonymous pen pal. At the same time, a zebra finch named Tucker soars into Manhattan to send a friend my shabby doodle of the Cool S.

These messages take hours or even days to send, depending on how far the bird has to fly, as that’s the point of Roost, the viral “slow-cial” app that’s making carrier pigeons cool again. Roost arrives at a time when people crave the opportunity to slow down and disconnect from the apps that constantly demand their attention and are embracing technology that adds friction.

“Everything on a phone is instantaneous these days — every single thing you do, it’s like you’re always getting some notification or something,” Roost creator Logan Mendelsohn told TechCrunch. “[Roost] is kind of a break from the instant. It’s resonating with people in a way where they don’t feel pressure all the time to have to do something.”

When you sign up for Roost, you choose four birds to add to your rookery, which allows you to send messages to your friends on the app.

Each bird moves at the speed that it travels in real life, so a falcon will deliver a message much faster than a hummingbird. (Yes, not every bird is a carrier pigeon, but including other species makes collecting birds and seeing your friends’ birds more interesting.) If you really want to slow things down, you can send snails or turtles instead.

A senior product manager in trust and safety at Ticketmaster, Mendelsohn started building Roost as a fun side project to use with his friends, but they loved the app so much that they encouraged him to publish it to the App Store.

Mendelsohn’s friends were onto something. The app developed a very small niche following, but it started to grow exponentially when a mother posted on Threads about how her daughter was communicating with her friends in Elizabethan English on an app that sends messages at the speed of actual birds.

Within three days after that post, the app grew from 10,000 to 100,000 users. Now, about five weeks later, Roost is about to hit 300,000 users.

“The people are what really make this platform, and what people kept talking about is how wholesome it is, and how whimsical it is, and how much this really helps them put more intention into what they’re saying to people,” Mendelsohn said. “There’s a lot less pressure when you know that the message isn’t going to someone immediately that I think has really resonated with the user base.”

As a trust and safety professional by day, Mendelsohn knows that any social platform — even his innocent bird app — has the potential to be abused. So, by default, only a user’s city is shared with their friends. You can choose to manually enable a “close friends” feature to share your precise location with specific people, however.

“I personally think that for any new platform that connects people, trust and safety should be the first thing they think about,” Mendelsohn said. “When you’re able to start at zero with that lens, then you can build it into the platform instead of doing it later.”

Privacy concerns were also front of mind when Mendelsohn created the “Pen Pals” feature, which allows you to exchange messages with anonymous users in your age group. When onboarding onto the feature, you are explicitly warned not to give out your actual contact information or personal details. The app deliberately does not support photo sharing yet, as Mendelsohn wants to build out more sophisticated content moderation tools first.

Given the sheer size and scope of Roost — did we mention there are mini games? — it doesn’t come as a surprise that Mendelsohn has used Claude Code throughout its development. But the kind of people flocking to Roost tend to be people who are fatigued by the state of the tech industry, which drove them to seek out a “slow-cial media” app in the first place.

X’s crowdsourced fact-checking system, Community Notes, will be updated to send users direct messages alerting them whenever a post they have interacted with has received a correction. The change, which is not yet live, was announced by X owner Elon Musk. He did not share a timeframe for its launch.

The update attempts to address one of the bigger criticisms about Community Notes — that corrections arrive too late to matter. A misleading post can accumulate views and reposts while its accuracy is disputed, and by the time it’s corrected, the damage has been done. By proactively notifying users when a post receives a correction, X is trying to extend the reach of the note beyond the original post. This could also allow users who spread false information to issue their own mea culpa, if they had been duped.

X’s Community Notes system was first established when the company was still known as Twitter, before Musk’s acquisition.

The idea was to introduce a different way to address misinformation on the platform, rather than require Twitter (now X) to be the centralized authority for moderation decisions. Instead, Community Notes contributors could suggest corrections and add critical details or missing information to posts. Consensus is achieved when people who rate the note as helpful are those who typically have different perspectives, and the note goes live.

A similar system has since been adopted by Meta as part of its broader moderation overhaul last year, which saw the company eliminate its partnerships with fact-checkers.

Though Community Notes makes sense for a company that wants to distance itself from the business of fact-checking, it’s also proven difficult to scale. A 2025 study of the feature by Spanish fact-checking site Maldita found that 85% of the proposed notes on X remain invisible to users, and only 8.3% get published and become visible. A separate study conducted by the Digital Democracy Institute of the Americas (DDIA), which encompassed 1.76 million notes published on X between January 2021 and March 2025, put the figure for unpublished notes even higher at 90%.

This weakens Community Notes as a system that surfaces information when it’s most needed, critics have pointed out. Plus, they’ve argued, people aren’t aware when a post they saw or boosted receives a correction later on, as there’s been no way to bring that information to their attention.

Musk’s proposal to send users alerts via X Chat (DMs) would address the latter issue, at least, assuming it goes live. X was asked for comment, but a response was not immediately available.