Grok 4.7 API: Pricing, 500K Context and Setup

Grok 4.7 is xAI's new flagship, released on 21 September 2026. The API model ID is grok-4.7, the context window is 500,000 tokens, and the list price is $2 per 1M input tokens and $6 per 1M output tokens.

That is the same headline price and the same speed as Grok 4.6. What changed is the model: a larger base, a longer reinforcement-learning run on multi-hour tasks, and stronger self-verification. It is live on RouterPlex today.

Grok 4.7 API pricing #

All figures are USD per 1M tokens. The first row is the standard rate for prompts under 200K tokens; the second is long-context pricing at or above 200K.

xAI lists cached input at $0.50 per 1M on this model ($1.00 at or above 200K). On RouterPlex, cached prompt tokens on grok-4.7 currently bill at the full input rate. The $0.50 cache-read tier is not passed through yet. Price a cache-heavy agent loop on $2 input until that row changes on the live grok-4.7 page.

A $5 prepaid balance buys 2.5 million input tokens, or 833,333 output tokens, if a workload used only one category. Real requests mix both, and reasoning tokens count as output.

The 200K cliff still applies #

xAI's long-context rule is unchanged, and it is not a marginal rate. Requests whose prompt reaches 200K tokens are billed at the higher rate for every token in the request.

199,000-token prompt on grok-4.7  →  199,000 × $2.00/1M  = $0.398
201,000-token prompt on grok-4.7  →  201,000 × $4.00/1M  = $0.804

Two thousand extra tokens double the cost of the request. If you are anywhere near 200K, trimming the prompt below the threshold is the highest-leverage cut available.

Grok 4.7 vs Grok 4.6 #

Headline price, context, and speed match. The upgrade is quality.

• xAI marks the 4.7 DeepSWE score as high effort.

xAI's line is that 4.7 uses a new, larger base than 4.6, trained with a longer RL run weighted toward problems that take many hours, and trained natively on the Grok Bot harness. xAI did not publish a parameter count. Treat the table above as vendor-run numbers on tasks xAI chose.

How Grok 4.7 compares on price #

Flagship list prices on RouterPlex, USD per 1M tokens:

On a 40K-in / 2K-out agent turn with no cache, Grok 4.7 costs $0.092. That is the same turn cost as Qwen3.8-Max, $0.250 for Claude Opus 5, $0.260 for GPT-5.6 Sol, and $0.500 for GPT-6 Astra or Claude Fable 5.

xAI's own launch table prices GPT-5.6 Sol Max at $4 / $20 and Fable 5.1 Max at $10 / $50. Those are vendor Max SKUs. The RouterPlex rows above are the catalog IDs you can actually send: gpt-5.6-sol at $5 / $30 and claude-fable-5 at $10 / $50.

xAI's OG line for 4.7 is "twice as fast, at half the price of comparable models." That comparison is against that Max-SKU class. Against Grok 4.6, 4.7 is the same sticker and the same served speed.

What xAI measured #

xAI published this table on the launch post. Vendor-run, on tasks xAI chose.

The pattern: 4.7 clears 4.6 on every row xAI showed. Fable 5.1 Max still leads CursorBench, Terminal-Bench, Briefcase, and HealthBench, at $10 / $50. Grok 4.7 is the value side of that frontier: close enough on several agent benches that the $2 / $6 sticker is the decision.

On safety, xAI says 4.7 ships a new safeguard stack, scores 62.4% on LatchBio biosafety, and allows 3.3% of risky dual-use prompts through on HackerBench v0.3.

Grok 4.7 Fast is a different SKU #

xAI also serves a Fast variant: twice the output speed at twice the price. That SKU is in Cursor and Grok Build. It is not on the public xAI API, so RouterPlex has no grok-4.7-fast (or similar) ID. If a Cursor screenshot shows Fast, that is Cursor's route, not grok-4.7 on api.routerplex.com.

The model you get on RouterPlex is the public API flagship: grok-4.7 at $2 / $6.

What the catalog actually lists #

Live RouterPlex row for grok-4.7:

Grok 4.7 is a reasoning model, and reasoning tokens are billed as output at $6 per 1M. A short reply is not always a cheap one. A two-token answer preceded by 200 reasoning tokens bills as 202 output tokens. Check usage.completion_tokens_details.reasoning_tokens if a simple prompt costs more than the visible completion suggests.

Call Grok 4.7 with RouterPlex #

Create a RouterPlex key, give it a hard budget, and send a standard chat-completions request:

curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.7",
    "messages": [
      {"role": "user", "content": "Name the riskiest assumption in this migration plan."}
    ]
  }'

The Python version uses the regular OpenAI client. Only the base URL, key, and model ID change:

import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["ROUTERPLEX_API_KEY"],
    base_url="https://api.routerplex.com/v1",
)
 
response = client.chat.completions.create(
    model="grok-4.7",
    messages=[{"role": "user", "content": "Design a rollback-safe deployment plan."}],
)
 
print(response.choices[0].message.content)

Claude Code and the Anthropic SDK reach the same model through the Anthropic-compatible format. Set ANTHROPIC_BASE_URL=https://api.routerplex.com and use grok-4.7 as the model. See the Claude Code setup guide for the full configuration.

Should you switch from Grok 4.6? #

• Coding agents and long-running tasks — this is the job xAI trained 4.7 for. Same sticker as 4.6, higher CursorBench, DeepSWE, EEBench, and Terminal-Bench scores on xAI's table. Switch and A/B on the same key.
• Legal and office-agent loops — Harvey and Briefcase both moved up versus 4.6. Still priced as a value flagship.
• Cache-heavy prefixes — price the loop on full $2 input until RouterPlex passes xAI's $0.50 cache-read tier. The 4.6 catalog row currently bills cache at full input as well, so the two IDs match on that line today.
• Anything near 200K tokens — the cliff is identical. Trim below the threshold, then pick a model.

Both are live on RouterPlex at xAI list prices with no per-token markup, so you can A/B them on the same prepaid key and compare real spend. See the live pricing for grok-4.7, the Grok 4.6 write-up, or the full Grok API pricing table for the rest of the xAI catalogue.