Musk's "price butcher" has arrived: Grok 4.5 takes on Opus, slashing prices to one-tenth
Title: Musk's "Price Slasher" Has Arrived: Grok 4.5 Takes on Opus at a Tenth of the Price
The biggest buzz in the AI world last night came from Musk.
SpaceXAI officially released Grok 4.5—the first major version update since the company went public a few weeks ago. Musk personally promoted it on X, directly positioning it against Anthropic's flagship model, Opus, claiming that this
💡 What You Will Learn
Title: Musk's "Price Slasher" Has Arrived: Grok 4.5 Takes on Opus at a Tenth of the Price The biggest buzz in the AI world last night came from Musk. SpaceXAI officially released Grok 4.5—the first
Headline: Musk's "Price Slasher" Arrives: Grok 4.5 Takes on Opus at a Tenth of the Price
The biggest news in the AI world last night came from Musk.
SpaceXAI officially released Grok 4.5 — the first major version update since the company went public a few weeks ago. Musk personally promoted it on X, directly targeting Anthropic's flagship model Opus, calling it "an Opus-level model, but faster, more efficient, and cheaper."
Big talk. But look at the pricing, and you'll see he's not joking.
This price list will keep competitors up at night
Grok 4.5's pricing: $2 per million input tokens, $6 per million output tokens.
Take a look at this comparison table.
| Model | Input (per million tokens) | Output (per million tokens) | Multiplier (Grok output = 1x) |
|---|---|---|---|
| Grok 4.5 | $2 | $6 | 1x |
| Opus 4.7 | $5 | $25 | ~4x |
| GPT-5.5 / 5.6 Sol | $5 | $30 | 5x |
| Fable 5 | $10 | $50 | ~8x |
Grok 4.5's output price is only a quarter of Opus's, an eighth of Fable 5's, and a fifth of GPT-5.5's.
The Decoder's take is blunt: "Grok 4.5 is so cheap that the benchmark gap probably doesn't even matter."
This isn't empty talk. SpaceXAI says Grok 4.5's token efficiency is twice that of other leading models — it consumes fewer tokens for the same task. Their real-world testing on SWE Bench Pro shows Grok 4.5 uses 4.2x fewer tokens than Opus 4.8 to complete tasks. Cheaper per token, fewer tokens per task — two cuts of the knife, and enterprise customers' spreadsheets will do the talking.
Does performance actually hold up? Let the data speak
Musk says Grok 4.5 is "roughly equivalent to Opus 4.7, but faster."
Looking at the public benchmark data, that's not all marketing.
On Terminal Bench 2.1, Grok 4.5 scores 83.3%, nearly tying GPT-5.5's 83.4%, and trailing Fable 5's 84.3% by just one percentage point. In coding scenarios, that gap is practically negligible.
The weak spot is DeepSWE 1.1 (the ability to solve real GitHub issues): Grok 4.5 at 53%, GPT-5.5 at 67%, Fable 5 at 70%. The gap is real. But on SWE Bench Pro (a harder set of engineering problems), Grok 4.5 hits 64.7%, even surpassing GPT-5.5's 58.6%.
An easy conclusion emerges: Grok 4.5 isn't the "strongest," but it's absolutely "strong enough." And at its price point, whether "the strongest" is worth paying five times more becomes a whole different question.
This was a long-planned ambush
The timing of Grok 4.5's release is extremely deliberate.
Tomorrow (Thursday), OpenAI is set to release GPT-5.6 Sol — touted as "the most powerful model ever." The Trump administration had previously restricted its release over security concerns, and now it's finally been cleared. OpenAI clearly wanted Sol to dominate the headlines.
Instead, Musk dropped Grok 4.5 a day early, welding the words "cost-performance" into the public discourse.
More critically, Grok 4.5 was co-trained with Cursor. SpaceX just acquired Cursor last month in a $60 billion stock deal — the company behind one of the hottest AI coding IDEs right now. This means Grok 4.5 was fed Cursor's coding-scenario data from the training stage, making it naturally suited for developer use cases.
It went live on Cursor, Grok Build, and the SpaceXAI console the same day. Plugins for Word, PowerPoint, and Excel also launched simultaneously.
No competing for benchmark crowns — it goes straight into your workflow.
The strategic logic is clear
Grok 4.5's pricing strategy isn't actually Musk's original idea. Chinese vendors have played this game for a while — Zhipu's GLM and DeepSeek did exactly the same: catch up on performance, then slash prices to the bone.
The difference is that Musk has the earnings pressure of a public company behind him, plus the ecosystem ambitions of the entire SpaceXAI. Grok 4.5 is just the first step. Its role isn't to crush GPT-5.5 or Fable 5, but to draw a line at the ceiling of "good enough" — above the line are others' premium prices, below it is Grok.
As for which side developers choose? Time will tell.
Written by our editorial team; tools listed here are tested or verified against public sources. Links point to official sites or GitHub repos for reference only — no paid placements.
