Last verified
RETIRED256K CONTEXTARCHIVE PRICEREDIRECTS TO GROK-4-3

Grok 4 API Pricing

Grok 4 is now an archive row: historical pricing was $3/M input, $0.75/M cached input, and $15/M output. SpaceXSI's May 15 retirement guide says retired slugs redirect to Grok 4.3 and are billed at current Grok 4.3 prices.

Input - per 1M tokens
$3.00/M
Historical Grok 4 archive
Output - per 1M tokens
$15.00/M
Archive launch price archive
Cached input
$0.75/M
Prompt cache discount
Effective - agentic blend
$2.26/M
92/8 split - 82% cache
§ 01 / TERMINAL

Grok 4 API cost calculator.

Calculator pre-loaded with Grok 4's archive rates. For live traffic after May 15, 2026, SpaceXSI (formerly xAI) says the retired slug redirects to Grok 4.3 and bills at Grok 4.3 pricing.

$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
Open full calculator (all models · share URL · CSV) →
§ 02 / SCENARIOS

Grok 4 cost on real workloads.

§ 03 / TOKENIZER

Grok 4 token counter and cost.

What's a token →
Estimate · grok-tokenizer-estimate · ≈3.85 chars/token Auto-counts as you type

This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.

Characters 249
Words 40
Tokens (estimated) 65 tokens
Cost as input · uncached $0.00019 USD
Cost as output · uncached $0.00097 USD
Cost as cached input $0.00005 USD
§ 04 / SHELF

Grok 4 vs similar models.

All models →
Model Input /M Output /M Effective blended Context Best for
Grok 4 Current $3.00 cache $0.75 $15.00 $2.26 agentic 92/8 256K Archive Grok 4 invoices
Grok 4.3 $1.25 cache $0.20 $2.50 $0.558 −4.1× 1M Current Grok default
Gemini 3.8 Flash $0.75 cache $0.075 $3.75 $0.481 −4.7× · promo 1M Google's pick for new projects
GPT-6.1 Sol $2.00 cache $0.10 $10.00 $1.29 −43% 1.05M OpenAI's balanced model for coding and agents
Claude Sonnet 5.5 $2.00 cache $0.10 $10.00 $1.29 −43% 1M Anthropic's current Sonnet
DeepSeek V4 Pro $1.32 cache $0.044 $3.96 $0.569 −4.0× 1M Low-cost reasoning alternative
§ 07 / BENCHMARKS

Grok 4 benchmarks

All benchmarks →

Grok 4 has published scores on 2 benchmarks we track; its best placing is #59 of 105 on ARC-AGI-2. All of these scores come from independent leaderboards, not from the lab.

Benchmark Score Rank Leader Source
ARC-AGI-2 Reasoning 16% ARC Prize verified, best result #59 of 105 GPT-6 Astra · 95% Official leaderboard ARC Prize Foundation
ARC-AGI-1 Reasoning 66.7% ARC Prize verified, best result #56 of 99 6 tied · 98.5% Official leaderboard ARC Prize Foundation

Grok 4 API pricing FAQ.

Grok 4 pricing questions, with SpaceXSI token rates separated from workload assumptions.

Q · 01 What was Grok 4's API price? +
The archive snapshot keeps Grok 4 at $3/M input, $0.75/M cached input, and $15/M output. Use these numbers for historical billing analysis, not new routing decisions.
Q · 02 Is Grok 4 still available? +
SpaceXSI's May 15, 2026 retirement guide says grok-4-0709 and related retired slugs redirect to grok-4.3. After the redirect, requests are billed at Grok 4.3's $1.25/M input and $2.50/M output pricing.
Q · 03 Does prompt caching apply? +
The archive row includes cached input at $0.75/M. The effective blend assumes 82% of input tokens hit cache; turn cache off in the calculator for fresh-input workloads.
Q · 04 Why keep an archive page? +
Teams still need old price rows to reconcile invoices and explain model migrations. This page explicitly marks the endpoint as retired so it does not imply new production availability.
Q · 05 What should replace Grok 4? +
Use grok-4.3. SpaceXSI states retired text slugs redirect to Grok 4.3, and the live pricing page lists Grok 4.3 at $1.25/M input, $0.20/M cached input, and $2.50/M output.
Q · 06 How accurate is the tokenizer estimate? +
The widget uses 4.875 characters per token as a planning estimate. Exact billing can vary with language, reasoning effort, cached prompts, and tool calls.