Grok Pricing 2026: xAI's Grok Cost Guide
xAI's Grok lineup in 2026 is three models deep: Grok 4.3 (high/medium) at $0.30/$1.50 per 1M tokens and Grok 4.1 Fast at $0.20/$0.50 with a 2M context window. Full breakdown, blended-cost math, and when each tier wins.
xAI's Grok lineup in mid-2026 is three models: Grok 4.3 (high) and Grok 4.3 (medium) at $0.30 / $1.50 per 1M tokens, and Grok 4.1 Fast at $0.20 / $0.50 per 1M with a 2,000,000-token context window. Both Grok 4.3 tiers share the same $0.66/1M blended cost; Grok 4.1 Fast is the cheap option at $0.29/1M blended. That's the whole pricing picture — the rest of this guide explains what each tier actually costs on real workloads and when to pick Grok over GPT, Claude, or Gemini.
Every number below is pulled live from our own pricing API (the same data the calculator pages use). Run your own workload through the cost calculator, or compare Grok head-to-head with other models on the model comparison page.
The Grok pricing table (mid-2026)
USD per 1M tokens, list rate. "Blended" uses a 70% input / 30% output weighting, matching typical chat and RAG traffic.
| Model | Input | Output | Blended (70/30) | Context |
|---|---|---|---|---|
| Grok 4.3 (high) | $0.30 | $1.50 | $0.66 | 1M |
| Grok 4.3 (medium) | $0.30 | $1.50 | $0.66 | 1M |
| Grok 4.1 Fast | $0.20 | $0.50 | $0.29 | 2M |
Source: llmcalculator.net live pricing API (/api/v1/models), snapshot taken 2026-07-16.
What each Grok tier costs on real workloads
List prices are abstract. Here's what 1,000 typical interactions cost on each Grok tier, using the same workload profiles the rest of the site uses:
| Workload | Grok 4.3 (high/medium) | Grok 4.1 Fast |
|---|---|---|
| Chat (1.5K in / 400 out) | $1.05 | $0.50 |
| RAG (8K in / 600 out) | $3.30 | $1.90 |
| Agent loop (12K in / 2K out) | $6.60 | $3.40 |
Workload math: 1000 × (in_tokens × input_price + out_tokens × output_price), prices from the table above. Chat example for Grok 4.3: 1000 × (1500 × $0.30 + 400 × $1.50) / 1,000,000 = $1.05.
Which Grok tier should you actually use?
Grok 4.3 (high) — the reasoning default
Use the "high" reasoning variant for hard reasoning, coding, and long-context work where you want Grok's strongest intelligence index (53 in our data). It's tagged for reasoning, coding, and long-context, with a 1M context window and a March 2026 knowledge cutoff.
Grok 4.3 (medium) — same price, lighter reasoning
Identical pricing to the high variant ($0.30/$1.50) but tuned for faster, lighter-weight reasoning (intelligence index 49). If your workload doesn't need the high-reasoning path, "medium" gives you the same cost with lower latency.
Grok 4.1 Fast — the cheap, long-context tier
At $0.20/$0.50 per 1M and a 2,000,000-token context window, Grok 4.1 Fast is the standout for bulk long-context work — full codebase or document-corpus inputs at roughly half the blended cost of Grok 4.3. Its intelligence index (43) is lower, so reserve it for retrieval, summarization, and classification rather than hard multi-step reasoning.
Grok vs the frontier competition
Grok's positioning in 2026 is "frontier-ish quality at deep-discount pricing." On blended cost it sits well below the flagship tier:
| Model | Blended $/1M | Context | Position |
|---|---|---|---|
| GPT-5.5 (high) | $3.88 | 922K | Frontier default |
| Claude Sonnet 5 | $6.60 | 1M | Agent / coding default |
| Gemini 3.5 Pro | $4.65 | 2M | Long-context frontier |
| DeepSeek V4 Pro (max) | $0.52 | 1M | Cheap reasoning |
| Grok 4.3 (high/medium) | $0.66 | 1M | Cheap + long-context |
| Grok 4.1 Fast | $0.29 | 2M | Cheapest 2M-context |
Source: llmcalculator.net live pricing API (/api/v1/models), snapshot 2026-07-16. Blended = 70% input + 30% output. See the full cross-provider breakdown in the 2026 LLM price comparison.
At $0.66/1M blended, Grok 4.3 is roughly 6× cheaper than GPT-5.5 ($3.88) and 10× cheaper than Claude Sonnet 5 ($6.60), with a 1M context window that matches Claude and beats GPT-5.5's 922K. The quality tradeoff is real — Grok 4.3's intelligence index (53) trails GPT-5.5 (59) and Sonnet 5 (61) — but for cost-sensitive long-context work, that gap is often worth the savings. The closest direct competitor is DeepSeek V4 Pro at $0.52/1M blended, which is cheaper still but offers only comparable-to-lower quality on most reasoning evals.
The bottom line
Grok in 2026 is a value play, not a flagship play. Default to Grok 4.1 Fast ($0.20/$0.50, 2M context) for bulk and long-context work where frontier-quality reasoning isn't required; reach for Grok 4.3 (high) ($0.30/$1.50) when you want Grok's best reasoning at still-low pricing. Pair either with a frontier tier like GPT-5.5 or Claude Sonnet 5 behind a cascading router for the hard 20% of requests — see how to cut your LLM bill by 60% for the routing logic.
By the LLMCalculator.net Research Team · Last updated: July 16, 2026
Related reading: 2026 LLM price comparison · Cheapest LLM API 2026 · How to cut your LLM bill by 60% · Claude vs GPT cost comparison · Live cheapest-LLM ranking.