Grok Mastery Pro ~7 min read Updated · August 13, 2026

Grok 4.5: the price shot heard round the labs.

On July 8, SpaceXAI launched Grok 4.5 — its strongest model, built for coding, agentic tasks, and knowledge work, and trained alongside Cursor. The headline isn't a benchmark crown. It's the invoice: $2 per million input tokens, $6 per million output, served at fast-model speeds. Here's what shipped, what SpaceXAI's own charts actually show, and where you can (and can't) use it today. Updated August 13, 2026: Grok 4.6 landed on August 12 at the same $2/$6 price — jump to section 04.

01 What launched

FactDetail
AnnouncedJuly 8, 2026, with public rollout continuing July 9
Built forCoding, agentic tasks, knowledge work — trained alongside Cursor
Price$2 in / $6 out per million tokens
Speed~80 tokens per second — flash-model speed for a frontier-class model
EfficiencyRoughly 2x the token efficiency of comparable leading models; on SWE Bench Pro tasks it used ~4.2x fewer output tokens than Claude Opus 4.8 (max)
WhereDefault model in Grok Build; available in Cursor (all plans) and via the SpaceXAI API

02 The honesty layer: read their own charts

SpaceXAI calls Grok 4.5 its smartest model — and Elon Musk calls it "Opus-class." Both can be true while the fine print stays interesting. On the benchmark charts in SpaceXAI's own announcement:

The honest read: Grok 4.5 is genuinely frontier-adjacent — trading places with the best, not lapping them. What's uncontested is the price-performance: models in its performance band typically cost 2–5x more per token. If your workload is high-volume coding or agent runs, that's the story that changes your bill.

03 Where you can use it today

Three doors, all open now — plus one closed one:

04 August 12, 2026: Grok 4.6 lands — same price

Five weeks after 4.5, SpaceXAI and Cursor released Grok 4.6 together. The pitch is narrower and more useful than "smartest model": it is built to stay on a job across many steps — long research runs, working across a codebase, turning a rough product idea into something that actually opens in a browser.

FactDetail
ReleasedAugust 12, 2026, jointly by SpaceXAI and Cursor
Built forLong-running agents; interactive and visual work
Price$2 in / $6 out per million tokens — unchanged from 4.5. A fast variant costs twice that.
Benchmark claimMatches GPT-5.6 Sol on the Artificial Analysis Intelligence Index (a composite of nine benchmarks)
WhereCursor, Grok Build, the SpaceXAI API, and partners including OpenRouter, Vercel and Cloudflare
Launch offer2x included usage in Cursor and Grok Build for the first week

Two things changed under the hood that you can actually feel. Cursor reports 4.6 does more self-checking on long runs — testing its own work before moving on — and produces stronger first passes on visual and interactive projects, establishing structure and visual language in one go instead of three.

The honest read: "matches GPT-5.6 Sol on a composite index" is a tie, not a crown — and it is a vendor-reported tie on a benchmark the vendor chose. The durable fact is the same one as July: the price did not move. A model in this performance band at $2/$6 is the story; the leaderboard position will change again before your invoice does.

Should you switch? If you are running agents that work for minutes rather than seconds, 4.6 is the one to test — that is precisely what it was trained for. For short one-shot prompts, 4.5 at the same price is unlikely to feel different, and the model ID you already have keeps working. See what Grok actually costs for how these rates sit against the subscription tiers.

Try it this week

  • If you use Grok Build or Cursor: check the model picker for 4.6 — the 2x-usage window ran for the first week from August 12
  • Re-run a task that a pricier model handles for you today, and compare output and token count
  • API users: the ~2x token efficiency means the real cost gap vs. list price is bigger than it looks — measure per completed task, not per token

This week's challenge

Take one real coding or document task you ran recently on Claude, GPT, or Gemini. Run the identical prompt through Grok 4.5 in Grok Build (free right now). Score three things: quality, speed, and — if you're on the API — total tokens consumed. You're not looking for a new religion; you're building your own price-performance table instead of trusting anyone's launch charts. Including SpaceXAI's. Including ours.

Up next in Grok Mastery

Grok's maker is now SpaceXAI

The February acquisition, the July rebrand, and what actually changed for users. Read the lesson →