xAI on July 8, 2026 launched its newest AI model, Grok 4.5, and founder Elon Musk called the model “an Opus-class model, but faster, more token-efficient and lower cost” than Anthropic’s flagship.
Quick Summary – TLDR:
- Grok 4.5 is priced at $2 per million input tokens and $6 per million output tokens, according to SpaceXAI.
- Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster, per Musk, though Anthropic’s Opus 4.7 costs $5 per million input tokens and $25 per million output tokens by comparison.
- On SWE Bench Pro, Grok 4.5 resolves tasks using 15,954 output tokens on average, about 4.2x fewer than Opus 4.8 (max) at 67,020.
- The model is served at 80 TPS and is now the default model in Grok Build, xAI’s command-line coding tool, and is available in Cursor on all plans, per SpaceXAI.
- xAI’s own benchmark charts show Grok 4.5 trailing Fable and GPT 5.5 on the DeepSWE 1.0 eval, at 62.0% versus Fable’s 66.1% and GPT 5.5’s 64.31%, even as it leads on other tests,.
What Happened?
xAI, now branded SpaceXAI, whose social platform X operates as a subsidiary of SpaceXAI, released Grok 4.5 as its newest flagship model for coding, agentic tasks, and knowledge work. The release is xAI’s first since the company went public several weeks earlier.
It is an Opus-class model, but faster, more token-efficient and lower cost, Musk wrote on X, positioning Grok 4.5 as a direct rival to Anthropic’s Opus line rather than a budget alternative.
The comparison is notable because xAI’s own announcement says Grok 4.5 excels at real engineering tasks and exceeds comparable leading models at these tasks, built on training data spanning coding, science, engineering, and math. The pricing gap is the sharpest part of the pitch: OpenAI’s priciest pricing tier, Sol, costs $5 for 1 million input tokens and $30 for 1 million output tokens, while its least expensive tier, Luna, costs $1 for 1 million input and $6 for 1 million output tokens, putting Grok 4.5’s output pricing on par with OpenAI’s budget tier, not its flagship.
Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency.https://t.co/i8HpU7w64k pic.twitter.com/oBjGtTsoNc
— SpaceXAI (@SpaceXAI) July 8, 2026
Benchmarks tell a mixed story
xAI’s published charts undercut the “Opus-class” framing on several counts. On the SWE Bench Pro resolve rate, Grok 4.5 scored 64.7%, behind Fable’s 80.4% and Opus 4.8’s 69.2%, though ahead of Opus 4.7 at 64.3%.
On Terminal Bench 2.1, Grok 4.5 scored 83.3%, essentially tied with GPT 5.5’s 83.4% and just behind Fable’s 84.3%. Grok 4.5’s clearest win came on the SWE Marathon resolution rate, where it scored 29.0% against Opus 4.8’s 26.0% and Fable’s 24.0%.
xAI’s argument, and the one Musk repeated on X, is that raw accuracy is not the whole picture. The company says Grok 4.5 was trained across tens of thousands of NVIDIA GB300 GPUs with a heavy reinforcement-learning investment in per-token intelligence, which xAI credits for the token-efficiency gap over Opus 4.8. The scale of that buildout mirrors the broader Nvidia’s trend of AI labs racing to secure GPU supply for ever-larger training runs.
A crowded release week
Grok 4.5 landed the day before OpenAI’s planned release of GPT 5.6, which is set to use its top Sol pricing tier and which OpenAI has called its “strongest model yet“.
That OpenAI release had previously been limited by the Trump administration over security concerns, adding a second regulatory subplot to the same week’s model race.
- Grok 4.5: xAI, launched July 8, 2026, priced at $2/$6 per million tokens.
- Opus 4.7 / Opus 4.8: Anthropic, priced at $5/$25 per million tokens for Opus 4.7.
- GPT 5.6 (Sol): OpenAI, priced at $5/$30 per million tokens, releasing the same week.
For coding-model buyers, the token-efficiency claim matters more than the headline price: a model that uses fewer output tokens per task can end up cheaper in practice even at a similar per-token rate, which is the argument xAI is making against Opus 4.8 specifically.
Grok 4.5 is not yet available in the EU in any xAI products or the API console, with EU availability expected in mid-July. That gap sits alongside xAI’s Word, PowerPoint, and Excel plugins, which the company says Grok 4.5 can use to build complex multi-sheet Excel models and design intuitive PowerPoint slide content.
SQ Magazine’s Takeaway
The “Opus-class” label is marketing shorthand more than a benchmark claim, though the $2/$6 per-million-token price tag is real and undercuts Anthropic’s Opus 4.7 by a wide margin. xAI’s own charts show Grok 4.5 splitting results with Anthropic’s and OpenAI’s models rather than beating them outright, and the strongest evidence for the comparison is cost and speed, not raw capability.
That distinction matters for teams choosing a coding model: Grok 4.5’s 4.2x token-efficiency edge over Opus 4.8 on SWE Bench Pro is the kind of operational detail that shows up in a monthly API bill long after a benchmark chart is forgotten, and it is the actual basis for Musk’s “faster, more token-efficient and lower cost” pitch.
