Grok

The job in the stack

The real-time-pulse seat: what's happening on X right now, plus a genuinely cheap fast API tier for high-volume tasks.

Pricing

Pricing verified 2026-10-14. Plans change — the source of truth is Grok's own pricing page.

Models & versions

• Grok 4.7 Current flagship, released September 21, 2026 — same $2/M input, $6/M output as 4.5 and 4.6, same 500K context and long-context billing cliff above 200K tokens. xAI describes a larger base model (reportedly 2.1T parameters, up from 4.6's 1.5T), a longer reinforcement-learning run weighted toward hard multi-step tasks, and better self-verification and long-context management. Multimodal (text + image input)
• Grok 4.6 Previous flagship, released August 12, 2026 — same $2/M input, $6/M output as 4.5, same 500K context, a 5-point jump on the Artificial Analysis Intelligence Index (61, level with GPT-5.6 Sol). Watch the long-context cliff: once a prompt crosses 200K tokens, xAI bills the whole request at $4/M input and $12/M output
• Grok 4.5 Earlier flagship (July 2026), Cursor-trained for coding and agentic work; same $2/$6 per M pricing. Still the tier most consumer plans actually serve
• Grok 4.1 Fast Speed/cost tier — $0.20/M input, $0.50/M output

The honest read

Questions operators ask

What changed

1. 2026-10-14 NEWS + verification: xAI launched The X Subscription on Oct 12, 2026 (x.ai/news/x-subscription; same tiers on cursor.com/pricing). Free, Plus $8, Premium $30 (~4x Plus), Super $100 (~15x), Ultra $200 (~50x), one weekly-reset usage pool shared across Grok apps, the Grok tab on X and Cursor; overage via top-ups at API list price. Premium bundles what SuperGrok + Cursor Pro + X Premium+ sold separately. Existing SuperGrok/X Premium plans continue unless the subscriber switches. The unconfirmed 'SuperGrok Plus $100' from Oct 10 was likely an early read of the Super tier. API rates unchanged.
2. 2026-10-10 Freshness sweep: no confirmed change since Sept 22. Grok 4.7 is still the flagship at $2/$6 per M; Grok 5 has no date. One Oct 8 third-party guide lists a 'SuperGrok Plus' at $100/mo that no other source confirms, so it was not recorded (logged for David).
3. 2026-09-22 NEWS SCAN: xAI shipped Grok 4.7 on September 21, 2026 — now the flagship, listed ahead of 4.6. Same $2/M input, $6/M output pricing and 500K context as the last two releases; xAI reports a larger base model (2.1T parameters vs. 4.6's 1.5T) and a longer RL run aimed at hard, multi-step tasks, plus better long-context handling. Still unresolved: which model the $30/mo SuperGrok consumer tier actually serves — this pass turned up sources claiming Grok 4.6, others still claiming 4.3 in staged rollout, and one citing a 128K context limit that contradicts every other source's 500K figure. x.ai's own pricing/model pages continue to block automated fetches. Left the SuperGrok tier note as-is rather than guess; logged for David again below.
4. 2026-09-11 Freshness sweep: consumer plan prices reconfirmed unchanged across multiple current sources — Free $0, X Premium $8, SuperGrok Lite $10, SuperGrok $30, X Premium+ $40, SuperGrok Heavy $300. Left the last item open rather than guess: sources disagree on which model SuperGrok's $30 tier is actually serving post-4.6 launch, one saying Grok 4.3 is still rolling out in stages, others implying 4.5. x.ai's own pricing page 403'd to automated fetch both attempts, so this couldn't be resolved directly — logged for David.
5. 2026-08-18 News scan: xAI shipped Grok 4.6 on August 12, 2026 — now the flagship, listed ahead of 4.5. The operator-relevant part is what did not change: still $2/M input and $6/M output, still a 500K context. What you get for the same money is about five points of measured intelligence (61 on the Artificial Analysis index, level with GPT-5.6 Sol), which makes Grok the cheapest model currently sitting at the frontier. Added the long-context billing cliff we had never documented: cross 200K tokens in a prompt and xAI reprices the entire request at $4/$12 per M, so a single oversized context window costs double. Dropped Grok 4.20 from the model list — two generations back and no longer a choice anyone should be making. Consumer plan prices unverified this pass and left as previously recorded.
6. 2026-07-27 Plan list corrected: the free tier and X Premium ($8) were collapsed into one row and X Premium+ ($40) was missing — both fixed. Grok 4.5 is reaching consumer plans in stages; SuperGrok Heavy is the only tier with full access so far. Verdict repriced against Claude Opus 5 (launched July 24 at $5/$25), which replaces Opus 4.8 as the comparison point — Grok 4.5 at $2/$6 still lands under half the cost.
7. 2026-07-09 NEWS SCAN: Grok 4.5 launched publicly today, trained with Cursor for coding/agentic work — Opus-class capability at $2/$6 per M tokens vs. Opus 4.8's $5/$25, using roughly 4x fewer tokens per task. Verdict and pros updated to reflect the price/capability shift; X-native real-time data remains the primary differentiator.
8. 2026-07-06 Added to the stack (real-time-pulse seat). Verified: SuperGrok Lite $10 (Mar 2026), SuperGrok $30, Heavy $300; API Grok 4.1 Fast $0.20/$0.50 per M.