In This Article
Key Takeaways
- SpaceXAI released Grok 4.5 on July 8, 2026, and opened it to the public the next day, positioning it for coding, agentic work, and routine knowledge tasks rather than as a consumer chatbot.
- Elon Musk described it as "an Opus-class model, but faster, more token-efficient and lower cost," and said internal assessment puts it "roughly comparable to Opus 4.7."
- API pricing is $2 per million input tokens and $6 per million output — well under Anthropic's Opus 4.7 at $5 / $25, per TechCrunch's comparison.
- It is available in Grok Build, in Cursor on all plans, and through the SpaceXAI console; TechCrunch noted its benchmark results looked "just short of best-in-class."
What SpaceXAI shipped
On Wednesday, July 8, 2026, SpaceXAI released Grok 4.5, and TechCrunch reported the model went public the following day. The framing is worth noting: SpaceXAI is not selling this as a chatbot. Per its own announcement, Grok 4.5 is built for "coding and app-building, office and clerical work, research, writing, and other forms of routine knowledge work." That is an agent-and-builder pitch, aimed at the same jobs Anthropic and OpenAI are chasing.
The model was trained alongside Cursor, the coding tool SpaceXAI now owns, and Axios first scooped the release the same day. It is available in Grok Build, inside Cursor on all plans, and through the SpaceXAI console.
The "Opus-class" claim, read carefully
The headline is a comparison. Elon Musk described Grok 4.5 as "an Opus-class model, but faster, more token-efficient and lower cost," and added that "our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster," per TechCrunch.
Two words in that sentence do a lot of work: "internal assessment." This is the vendor's own read, not an independent benchmark. TechCrunch noted that Grok 4.5's published results looked "just short of best-in-class," and did not independently verify the comparison. The honest way to treat any "as good as Opus" claim is the same as we treated Anthropic's own Opus 4.8 positioning: a reason to run your own evals, not a reason to assume.
The price is the actual story
Where the release gets concrete is cost. Grok 4.5 runs at $2 per million input tokens and $6 per million output tokens. For contrast, TechCrunch put Anthropic's Opus 4.7 at $5 input / $25 output — so on output tokens, the part that dominates most agent bills, Grok 4.5 is roughly four times cheaper.
That is the same pattern we keep seeing in 2026: the interesting number is not the benchmark, it is the cost of doing the same job. A model that is "roughly comparable" at a quarter of the output price changes the math on any workload that burns tokens by the millions — batch jobs, long agent runs, high-volume extraction. It does not change the math on a task where a two-point accuracy gap costs you a customer.
Where it fits in a crowded July
Grok 4.5 did not arrive in a quiet week. It followed Anthropic's Claude Sonnet 5 becoming the default at the end of June and landed alongside OpenAI's GPT-5.6 family, all within days of each other. The common thread across every one of these releases is downward price pressure, not a new capability nobody had before.
SpaceXAI says it trained Grok 4.5 across tens of thousands of NVIDIA GB300 GPUs, per its announcement — a reminder that the price war is being funded by enormous compute build-outs, a dynamic we unpacked in the 2026 compute boom. For a working builder, the takeaway is narrow and useful: if you have agent or batch workloads parked on a pricier model, Grok 4.5 is now worth a head-to-head eval on your own data before you commit another month of spend.
Sources: TechCrunch — SpaceXAI releases Grok 4.5; SpaceXAI — Introducing Grok 4.5 (primary announcement); Axios — Scoop: Musk's SpaceXAI releases Grok 4.5; Bloomberg — SpaceXAI, Cursor launch Grok 4.5. Analysis and framing by Precision AI Academy.
Common questions
When did Grok 4.5 launch? SpaceXAI announced Grok 4.5 on Wednesday, July 8, 2026, and made it publicly available the next day, per TechCrunch and Axios.
How much does Grok 4.5 cost? API pricing is $2 per million input tokens and $6 per million output tokens, per SpaceXAI's announcement — below Anthropic's Opus 4.7 at $5 input / $25 output.
Is Grok 4.5 really as good as Claude Opus? Elon Musk called it "Opus-class" and "roughly comparable to Opus 4.7," but that is SpaceXAI's internal assessment, not an independent benchmark. TechCrunch described its results as "just short of best-in-class." Treat it as a reason to test on your own workload.
Where can I use Grok 4.5? It is available in Grok Build, inside Cursor on all plans, and through the SpaceXAI console.