Grok 4.7 lists at $2 per million input tokens and $6 per million output. GPT-6.1 Sol lists at $2 and $10. That 40% gap on output is the headline reason developers look at xAI, and it holds down the ladder: Grok 4.3 at $1.25 and $2.50 with a 1 million token context against GPT-6 Luna at $0.10 and $0.50. The models aren’t interchangeable, but the price sheets, checked on both companies’ own pages on November 2, 2026, are easy to line up.
The structural differences matter more than the token prices for a team planning spend. xAI tiers rate limits by cumulative spend since January 1, 2026 and starts every account at 150 requests per second on Grok 4.7. OpenAI tiers by credit purchases and starts at a $500 monthly cap. xAI also lets prepaid credits roll into postpaid invoicing, while OpenAI stays prepaid. Neither gives a new account a signup credit.
No free credit on either side
Neither company publishes a credit for new accounts. xAI’s quickstart tells you to sign up at console.x.ai and load the account with credits before calling the API, and the default invoiced limit is $0, meaning only prepaid credits are spent until you raise it. Third-party sites still describe xAI signup bonuses and a data-sharing credit, but xAI doesn’t publish either, so we leave those figures out.
OpenAI is the same story. Its rate limit docs show a Free tier, but that’s a $100 monthly usage ceiling for allowed countries, and the first paid tier opens after $5 in credit purchases. The only no-cost items are a free moderation model and 1 GB of file search storage. Budget a small purchase on either platform to test, and treat anything that shows up in a console as a promotion rather than a policy.
Token prices model by model
xAI lists Grok 4.7, 4.6 and 4.5 at $2 input and $6 output per million tokens with a 500,000 token context, and Grok 4.3 and the 4.20 variants at $1.25 and $2.50 with a 1 million token context. Grok Build 0.1 is $1 and $2. Cached input is $0.50 on 4.7 and $0.20 on 4.3. Any request whose prompt reaches 200,000 tokens is billed at double rates for all tokens in it, so a long document can double the cost of a request without warning.
OpenAI lists GPT-6 Astra at $10 and $50, GPT-6.1 Sol at $2 and $10, GPT-6.1 Terra at $1 and $5 and GPT-6 Luna at $0.10 and $0.50, with cached input at one twentieth of list on the 6.1 models. The sub-dollar tier is where OpenAI pulls ahead; xAI’s cheapest text model still costs $2 per million output, four times Luna.
OpenAI also publishes a 50% discount for the Batch API and prices Flex processing at batch rates. xAI publishes no batch pricing, so asynchronous bulk work has a clear home. xAI does list image generation at $0.02 to $0.05 per image and video at $0.02 to $0.08 per second, which OpenAI prices separately.
How much traffic you get on day one
xAI sets tiers by cumulative API spend since January 1, 2026: Tier 0 at $0, Tier 1 at $50, Tier 2 at $250, Tier 3 at $1,000 and Tier 4 at $5,000, with enterprise on request. Tiers never downgrade. Tier 0 already allows 150 requests per second and 50 million tokens per minute on Grok 4.7, and 37 requests per second with 10 million tokens per minute on Grok 4.3. Limits are enforced per second, derived from a per-minute figure divided by 48. That’s production headroom before you’ve spent a dollar.
OpenAI’s Build tier, reached with $5 in purchases, allows 5,000 requests and 1 million tokens per minute on the Sol family and 2 million tokens per minute on Luna, under a $500 monthly cap. Launch at $100 raises the cap to $5,000, and Grow at $500 raises it to $200,000 with 40 million tokens per minute on Sol. Batch jobs use a separate queue limit and don’t consume those numbers.
Prepaid, postpaid, or both
xAI runs a hybrid. Prepaid credits are always consumed before postpaid usage accrues, and a monthly soft spending limit governs how far postpaid can go, with the API stopping when both are exhausted. Setting that limit to 0 keeps a team on prepaid only, which is the default. Cards, US bank accounts and Link are accepted. For a team that hates topping up, this is the feature that matters.
OpenAI is prepaid throughout the self-serve tiers, since each tier is defined by total credit purchases. That protects you from surprise invoices but means a Grow tier organization with a $200,000 cap still has to keep a balance funded. Both consoles expose per-team or per-organization spend controls.
What happens to your prompts
xAI stores all API requests and responses encrypted at rest for 30 days by default and then deletes them, doesn’t use that stored data for training, and never trains on API inputs or outputs without explicit permission. Zero data retention is available but disables features that rely on stored data. xAI’s documentation now carries the SpaceXSI name after the corporate change, so don’t be thrown by it.
OpenAI keeps abuse monitoring logs for up to 30 days, doesn’t use API data for training unless you opt in, and offers Zero Data Retention or Modified Abuse Monitoring to approved customers. Some endpoints such as files and batches store state until you delete it regardless.
Startup programs taking applications
Grok for Startups is live at x.ai/startups with an application form. The program covers Grok, Grok Build, Grok Bot and the Grok API from a single credit balance, asks for venture backing, a live product and a company email matching your website, and sizes grants to funding and current AI spend without publishing a dollar figure. Partner VCs and accelerators can offer a guaranteed minimum to portfolio companies.
OpenAI for Startups requires a unique referral code from a partner VC and lists free API credits, rate limit upgrades and solutions engineer time as possible benefits, again without a published amount. Credits revert to standard billing once spent. Neither program is open to bootstrapped founders, so if you haven’t raised, the prepaid route is the only route.
The winner: Tie
xAI is cheaper on flagship output, more generous on starting rate limits and currently accepting startup applications. OpenAI is cheaper on cached input and budget models, has a published batch discount, and offers a clearer path from $5 to a $200,000 cap. Pick by where your tokens land, not by brand.
Choose xAI if output-heavy generation dominates your bill, you want 150 requests per second from day one, or you need a 1 million token context at $1.25 input.
xAI Grok API offers (4)Choose OpenAI if you run batch or background jobs, rely on a sub-dollar budget model, or want cached input at one twentieth of list price.
OpenAI API offers (5)