Grok Build has two pricing paths: $99/mo flat rate (SuperGrok) or pay-per-token with an API key. This guide breaks down both options, compares them to Claude Code, and helps you pick the cheapest path for your usage level.
Grok Build pricing options
Option 1: SuperGrok subscription ($99/mo)
- Unlimited Grok Build CLI usage
- All modes (code, plan, ask)
- Multi-agent parallel subagents included
- Skills/Plugins marketplace access
- MCP server support
- 256K context window
- No per-token charges
- Browser OAuth authentication
Best for: Daily users, heavy coding sessions, teams that don’t want to track token spend.
Option 2: API key (pay-per-token)
| Cost component | Price |
|---|---|
| Input tokens | $1.00 / 1M tokens |
| Output tokens | $3.00 / 1M tokens |
| Via OpenRouter | Same rates, aggregated billing |
- No monthly commitment
- Same features as SuperGrok
- Set
XAI_API_KEYenvironment variable - Works with OpenRouter for unified billing across providers
Best for: Light users, CI/CD pipelines, developers who use multiple tools and want one bill.
When does SuperGrok break even?
The break-even calculation depends on your daily token usage:
| Daily usage | Monthly tokens | API cost/mo | Cheaper option |
|---|---|---|---|
| Light (50K in + 20K out) | 1.5M in + 600K out | ~$3.30 | API key |
| Moderate (200K in + 80K out) | 6M in + 2.4M out | ~$13.20 | API key |
| Heavy (500K in + 200K out) | 15M in + 6M out | ~$33.00 | API key |
| Power (1M in + 400K out) | 30M in + 12M out | ~$66.00 | API key |
| Extreme (2M in + 800K out) | 60M in + 24M out | ~$132.00 | SuperGrok |
Break-even point: roughly 45M input + 18M output tokens per month. That’s about 6-8 hours of active coding per day with constant agent interaction.
If you use Grok Build as your primary tool all day, SuperGrok saves money. If you use it a few times per week, the API key is significantly cheaper.
Grok Build vs Claude Code pricing
| Feature | Grok Build (SuperGrok) | Grok Build (API) | Claude Code (Max 5x) | Claude Code (API) |
|---|---|---|---|---|
| Monthly cost | $99 | Pay-per-use | $100 (Sonnet) / $200 (Opus) | Pay-per-use |
| Token limits | Unlimited | None (pay what you use) | Usage-based with caps | None |
| Input cost | Included | $1/1M | Included (with limits) | $3/1M (Sonnet) |
| Output cost | Included | $3/1M | Included (with limits) | $15/1M (Sonnet) |
| Context window | 256K | 256K | 200K | 200K |
| Throttling | None reported | None | Yes, at high usage | None |
Cost comparison scenarios
Scenario 1: Solo developer, moderate use (2-3 hours/day)
Estimated monthly: 8M input tokens, 3M output tokens
| Option | Monthly cost |
|---|---|
| Grok Build API | $17.00 |
| Grok Build SuperGrok | $99.00 |
| Claude Code API (Sonnet) | $69.00 |
| Claude Code Max 5x | $100.00 |
Winner: Grok Build API key at $17/mo.
Scenario 2: Full-time power user (6+ hours/day)
Estimated monthly: 50M input tokens, 20M output tokens
| Option | Monthly cost |
|---|---|
| Grok Build API | $110.00 |
| Grok Build SuperGrok | $99.00 |
| Claude Code API (Sonnet) | $450.00 |
| Claude Code Max 5x | $100-200 (throttled) |
Winner: Grok Build SuperGrok at $99/mo with no throttling.
Scenario 3: CI/CD pipeline (automated, bursty)
Estimated monthly: 5M input tokens, 2M output tokens (concentrated in short bursts)
| Option | Monthly cost |
|---|---|
| Grok Build API | $11.00 |
| Grok Build SuperGrok | $99.00 |
| Claude Code API (Sonnet) | $45.00 |
Winner: Grok Build API key at $11/mo.
Scenario 4: Team of 5 developers
Each using ~15M input, 6M output per month.
| Option | Monthly cost (team) |
|---|---|
| Grok Build API (shared key) | $165.00 |
| Grok Build SuperGrok (5 seats) | $495.00 |
| Claude Code Max 5x (5 seats) | $500-1000 |
Winner: Grok Build API with shared key at $165/mo. SuperGrok only wins if each developer is a power user hitting 45M+ input tokens individually.
Hidden costs to consider
Context window usage. Grok Build’s 256K context means fewer /compact cycles and less repeated context injection. This reduces total token consumption compared to tools with smaller windows.
Multi-agent overhead. Grok Build’s parallel subagents consume tokens for coordination. A complex refactor might use 20-30% more tokens than a single-agent approach, but completes faster.
Skills/Plugins. Marketplace skills are free to install. They don’t add token costs beyond the prompts they inject.
MCP servers. MCP tool calls consume tokens for the request/response. Heavy MCP usage (database queries, file operations) adds to your token count on the API plan.
How to track your spending
# During a session
/cost
# After session
grok --last-cost
# Set a budget alert (API key mode)
export GROK_BUDGET_ALERT=50 # Warns at $50/moRecommendation
| Your situation | Best option |
|---|---|
| Try it out, light use | API key |
| Primary daily tool, heavy use | SuperGrok ($99/mo) |
| CI/CD automation only | API key |
| Team, mixed usage | API key with shared billing |
| Team, everyone uses it all day | SuperGrok per seat |
| Already on Claude Code Max, want to compare | API key (run both, compare) |
For the full feature breakdown, see the Grok Build complete guide. For a head-to-head feature comparison, see Grok Build vs Claude Code.
FAQ
Is SuperGrok really unlimited? No fair-use caps?
As of launch (May 2026), xAI has not published any fair-use limits on SuperGrok CLI usage. There’s no throttling reported by early users. This could change, but right now it’s genuinely unlimited.
Can I use OpenRouter to pay for Grok Build?
Yes. Set your OpenRouter API key and configure Grok Build to route through OpenRouter. You get unified billing across all your AI tools. The token rates are the same ($1/1M input, $3/1M output).
Does the $99/mo include Arena Mode?
Arena Mode is listed as “coming soon.” xAI hasn’t confirmed whether it’s included in SuperGrok or will be an add-on. Current expectation is that it’s included.
Can I switch between SuperGrok and API key?
Yes. If you cancel SuperGrok, Grok Build falls back to API key authentication. Set XAI_API_KEY and you keep working. No data loss, no config changes needed.
How does Grok Build’s pricing compare to Codex CLI?
Codex CLI uses OpenAI models. At GPT-4 rates ($10/1M input, $30/1M output), it’s significantly more expensive per token than Grok Build’s API pricing. For equivalent usage, Grok Build API is roughly 10x cheaper on input tokens.
Is there a free tier?
No free tier for Grok Build. The cheapest entry point is the API key with pay-per-token. A light session might cost $0.10-0.50, so you can try it for very little money.