SpaceXAI released Grok 4.5 on July 8, 2026 — the company's first model since its IPO several weeks earlier. Elon Musk compared the new release to Anthropic's Opus class, describing it as roughly comparable to Claude Opus 4.7 but "much faster" and meaningfully cheaper, setting up a direct contest at the top of the market.
Key takeaways
- Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens
- Claude Opus 4.7 costs $5/$25 — 2.5× more expensive on input, over 4× more on output
- SpaceXAI claims 'twice greater token efficiency' versus competing models
- Grok 4.5 is SpaceXAI's first model launch since its public market debut
- Musk: Grok 4.5 is "roughly comparable to Opus 4.7, but much faster"
Pricing position and model architecture
Grok 4.5 is positioned as a general-purpose workhorse covering coding and app development, office and clerical tasks, research, writing, and other forms of routine knowledge work. SpaceXAI is not pitching this as a narrow specialist but as an all-round replacement for the type of work that today flows through Opus-class models at significantly higher cost.
The pricing arithmetic is straightforward. At $2 per million input tokens and $6 per million output tokens, Grok 4.5 undercuts Anthropic's Claude Opus 4.7 ($5/$25) by 2.5× on input and more than 4× on output. OpenAI's tier structure runs from Luna ($1/$6) at the low end to Sol ($5/$30) at the top. Grok 4.5 sits in the mid-premium bracket — above Luna in price, but well below Opus and Sol.
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Grok 4.5 | $2 | $6 |
| Claude Opus 4.7 | $5 | $25 |
| OpenAI Luna | $1 | $6 |
| OpenAI Sol | $5 | $30 |
SpaceXAI's claim of "twice greater token efficiency" versus competitors is self-reported and has not yet been independently verified. If it holds in production workloads, the real-world savings would exceed what the nominal price differential implies: a model that accomplishes the same task in half the tokens effectively costs far less, regardless of per-token list price. For enterprises running large-scale AI workloads, that difference compounds quickly across millions of API calls per day.
Benchmarks and competitive positioning
SpaceXAI published aggregate benchmark metrics on Wednesday suggesting Grok 4.5 is competitive with other leading models — though not quite at the top of every leaderboard. The company released a summary chart but did not publish granular results for individual benchmarks in its press materials.
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.
Elon Musk, founder of SpaceXAI, posting on X
Musk added more precision in a follow-up post: "Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster." That pairing — comparable capability at lower latency and cost — is the core product thesis. Caveats apply, however: both the token-efficiency claim and the benchmark comparisons come from SpaceXAI itself, not independent evaluators. Self-reported benchmarks in the AI industry have a well-documented history of selective framing.
The week of July 8, 2026 was unusually dense for model releases. OpenAI simultaneously announced GPT 5.6 — a model previously blocked by the Trump administration over national security concerns — describing it as its "strongest model yet". Two premium releases in the same week will produce the first direct market comparison between the two.
Why it matters
Grok 4.5 is the first concrete test of SpaceXAI's ability to compete as a standalone public company. Previously, xAI operated in the shadow of X and Musk's personal brand. IPO investors demand hard results: pricing aggression is the first tangible move, and it is legible.
The price pressure on the premium segment is measurable. Enterprises that run Opus-class models for coding, document analysis, or knowledge automation could cut token costs by 60–80% if Grok 4.5 delivers on its stated capabilities. That is a large enough differential for CTOs and CFOs to run a serious evaluation, even at companies that have historically preferred Anthropic or OpenAI on reliability and safety track records.
SpaceXAI's ownership of its own compute infrastructure — the Colossus cluster — gives it a structural advantage in pricing that cloud-dependent competitors lack. If the company can sustain these rates at scale while demonstrating the claimed efficiency gains, it will force the entire market to revisit margin assumptions.
What's next?
- SpaceXAI opened public access to Grok 4.5 on Thursday, July 9, following the closed beta program
- Independent token-efficiency benchmarks from the developer community will appear within days — production workload results will be the real test of SpaceXAI's claims
- OpenAI's GPT 5.6 launch in the same week creates the first head-to-head market comparison between the two premium contenders
