/ Cheaper And faster coding AI arrives today.
Mike Alvarez Joined TheTweaks as an emerging tech , startup and launches journalist. He comes up with those things or products which are not there yet and are hardly possible to come close to. He covers the latest technology, startups and new product launches and releases. He worked 4 years in a startup company in Austin and later he joined TheTweaks and he is proving himself here with his commendable writing performance. He is also the publisher of a small consumer robotics newsletter.
SpaceXAI released Grok 4.5 on July 8, 2026 and the model’s marketing is quite unique compared to the rest of the AI industry’s race for leaderboard superiority: it’s not selling itself as the smartest. It’s selling itself as the best deal.
Grok 4.5 is the company’s first dedicated coding and autonomous agent model and it’s also the first real output of SpaceXAI’s recent $60 billion acquisition of Cursor, the AI coding tool that struck its deal with the company just a couple of weeks ago.
The team behind Cursor confirmed that they worked together with SpaceXAI on training Grok 4.5 and the model is now integrated as the default option for all Cursor plans, Grok Build and SpaceXAI API console. For the time being, Elon Musk’s company is providing free access to Grok 4.5 in Grok Build and Cursor for a limited time period but it’s not yet available to EU users within the regional release that is expected by mid July.
Grok 4.5 is available right now, starting from July 8–9, 2026, within Grok Build, all Cursor plans and the SpaceXAI grok API console. Setting it up will take just an API key and a few lines of code. Free usage will be provided for a limited time period in Grok Build and Cursor until regular pricing starts.
All models launched in 2026 come with a benchmark chart but in the case of Grok 4.5 the more interesting story is about what SpaceXAI is betting the company on. Instead of trying to beat every leaderboard, Musk positioned Grok 4.5 as an economical model one that lands close to the top tier on capabilities but undercuts all competitors heavily on price and speed. Musk called the model “Opus-class” but faster and cheaper and added that his internal assessment of Grok 4.5 puts it on par with Claude’s previous Opus 4.7 but noticeably faster.
The pricing confirms this positioning. Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens. This is less than half of what Anthropic charges for Claude Opus 4.8, which costs $5 per million input and $25 per million output tokens. This pricing also undercuts all the newly released GPT 5.6 family models except the cheapest tier of OpenAI’s Luna that matches Grok 4.5 pricing for input costs.
SpaceXAI claims that the model operates at the speed of about 80 tokens per second and achieves near-double token efficiency of other similar models. This means that Grok 4.5 might need less tokens to provide the same result thus further increasing the price advantage in practice.
On coding specific benchmarks released by SpaceXAI, Grok 4.5 does not lead the group but lands somewhere in the middle of the pack, competing with Anthropic’s latest “Fable” model and OpenAI’s GPT-5.5 while outperforming Claude’s Opus 4.7 on all except the last test. The only exception is the SWE Marathon test that represents long horizon task resolution; the pass rate of Grok 4.5 was higher compared to Opus 4.8.
On general coding evaluation tests, like SWE Bench Pro and Terminal Bench,Grok 4.5 pushes back to both Fable and GPT-5.5 but stays close to Opus 4.8. Verification of these numbers is rather difficult because SpaceXAI conducted them using its own harness setup which means that these numbers should be taken with a grain of salt since they were provided by the vendor.
The reason why Grok 4.5 is notable is not only the model itself but also the company that launched it. SpaceXAI is the rebranded version of what used to be xAI, now incorporated in SpaceX after its IPO this year. This means that Musk is developing rockets, satellites and frontier AI models under the same roof, with the public pitch to investors relying on a market potential of almost $28 trillion, most of it being related to AI.
Grok 4.5 is the first real life test of that story and the timing is important. The reports suggest that Cursor lost its share of the AI coding tools market from roughly 41% a year ago to 26% by this spring, thanks to Claude’s advancement with developers. A cheap, fast and purpose built coding model is the easiest way to win that share back.
It’s quite surprising that Grok 4.5 has been trained using the same GPU capacity that SpaceXAI rents to its competitors, including Anthropic and Google. As the ambition of SpaceXAI increases, the question is whether it will continue renting its GPUs to competitors or start using it for its own training. This will affect future releases of Grok.
Grok 4.5 is not trying to win the intelligence battle and this is exactly what makes it so interesting. SpaceXAI is making a structural bet on the fact that most of enterprise AI spend won’t go to the most intelligent model but to the one that is cheap, fast and capable enough to be used at scale in the coding pipeline. On raw intelligence Grok 4.5 lags behind Anthropic’s Fable and OpenAI’s GPT-5.5 but its price per output token is less than half of the price of Claude Opus 4.8, which means that it doesn’t need to win on intelligence.
For teams working with heavy volumes of agentic coding workload, when token prices accumulate very quickly, this economics might become much more important than a few benchmark points. The question is whether SpaceXAI can maintain such pricing once it stops renting its GPU capacity to companies like Anthropic and Google and starts using it for itself.
Tim Cook’s 15 years as Apple’s chief executive officially ended on September 1 and now the torch has been passed on to John Ternus, who is the company’s long-time hardware. It’s Apple’s first leadership change…
.jpg&oldid=1139730095){kind=link}
Be respectful and constructive. Have a question or feedback? We’d love to hear from you. Contact us at contact@thetweaks.com