News Article · Jul 8, 2026 at 11:45 PM
2 min read 0
Member
SpaceXAI Launches Grok 4.5, Undercuts Rivals on Price With $2 per Million Input Tokens
Industry #Cursor #benchmarks #SpaceXAI #Grok 4.5 #AI pricing #coding models

SpaceXAI Launches Grok 4.5, Undercuts Rivals on Price With $2 per Million Input Tokens

SpaceXAI released Grok 4.5, its first model since going public and acquiring Cursor. Priced at $2 per million input tokens, it undercuts OpenAI and Anthropic while targeting coding and agentic tasks. EU availability is expected in mid-July.

SpaceXAI released Grok 4.5 on July 8, 2026, its first model since the company went public and acquired AI coding startup Cursor for $60 billion in stock. The model is designed for coding, agentic tasks, and knowledge work, and is available through Grok Build, Cursor, and the SpaceXAI console.

Grok 4.5 costs $2 per million input tokens and $6 per million output tokens. That is 80 percent cheaper than Anthropic's Fable 5, which charges $10 input and $50 output per million tokens, and 60 percent cheaper than OpenAI's GPT-5.5 at $5 input and $30 output per million tokens.

Benchmark performance shows mixed results

On Terminal Bench 2.1, which tests complex command-line tasks, Grok 4.5 scored 83.3 percent, nearly matching GPT-5.5 at 83.4 percent and trailing Fable 5 at 84.3 percent by just one point. But on DeepSWE 1.1, which measures the ability to resolve real GitHub issues, Grok 4.5 hit 53 percent, well behind GPT-5.5 at 67 percent and Fable 5 at 70 percent. On SWE Bench Pro, a curated set of harder software engineering problems, it scored 64.7 percent, beating Opus 4.8 in some configurations but falling short of Fable 5's 80.4 percent.

  • Grok 4.5 uses 4.2 times fewer tokens than Opus 4.8 on SWE Bench Pro tasks, according to xAI.
  • The model delivers results at 80 tokens per second.
  • Training relied on tens of thousands of Nvidia GB300 GPUs with heavy data filtering and deduplication.
  • The reinforcement learning stage covered hundreds of thousands of tasks, mostly from software engineering, with automated scoring.
  • Plugins are live for Word, PowerPoint, and Excel.

Pricing strategy echoes Chinese AI vendors

The pricing approach mirrors what Chinese companies like Zhipu and DeepSeek have done: get close enough on performance, then win on price. Lower per-token pricing and fewer tokens per task make Grok 4.5 by far the cheapest option in this performance tier, assuming the gains hold up in practice. The model is not available in the European Union yet, with SpaceXAI targeting a mid-July launch. Elon Musk described Grok 4.5 as an "Opus-class" model aimed squarely at coding and agentic work, not casual chat. The release marks SpaceXAI's first joint model with Cursor since the acquisition closed in mid-June.

Fact check

  • Grok 4.5 costs $2 per million input tokens and $6 per million output tokens.

    verified · source

  • On Terminal Bench 2.1, Grok 4.5 scored 83.3 percent, nearly matching GPT-5.5 at 83.4 percent and trailing Fable 5 at 84.3 percent.

    verified · source

  • SpaceXAI acquired Cursor for $60 billion in stock in mid-June 2026.

    reported · source

  • Grok 4.5 is not available in the EU yet, with a target launch of mid-July.

    reported · source

Source reporting (6)

0 Comments

No comments yet

Be the first to share your thoughts on this article.

Join the conversation

You need to be registered and logged in to comment on blog articles.

Who Is Online

In total there are 50 users online: 0 registered, 45 guests and 5 bots.

Most users ever online was 4,502 on 28 Jun 2026, 10:02 am.

Bots: Applebot Bingbot Other Bot PetalBot SemrushBot

Users active in the past 15 minutes. Total registered members: 367