Google’s Gemini 3.8 Flash has the lowest input price of these until 31 December, then it doubles. On output, Grok 4.3 is cheaper. Here are the published per-token prices for Gemini and Grok, read off the vendors’ own pricing pages on 16 September 2026 — and what Meta does and does not publish.
- Gemini 3.8 Flash: $0.75 in, $3.75 out per million tokens — “through December 31, 2026”.Google, Gemini Developer API pricing, read in a browser 16 Sep 2026 (the docs refuse automated fetches): input “$0.75 through December 31, 2026.$1.50 starting January 1, 2027.”
- It doubles on 1 January 2027 to $1.50 and $7.50. That is on the page today, not a forecast.Google, Gemini Developer API pricing, read in a browser 16 Sep 2026: output “$3.75 through December 31, 2026.$7.50 starting January 1, 2027.”
- Grok 4.6: $2.00 in, $6.00 out under 200k tokens, doubling above that threshold.xAI, Pricing, read at source 17 Sep 2026: grok-4.6 “$2.00” input and “$6.00” output below 200k prompt tokens, “$4.00” and “$12.00” at or above it.
- Grok 4.3 is the cheap one — $1.25 and $2.50, with a 1M context.xAI, Pricing, read at source 16 Sep 2026: grok-4.3 “$1.25” input and “$2.50” output per 1M tokens.
- Google has a free tier and xAI’s page lists none. The free tier costs you something else: Google’s row for “Used to improve our products” reads Yes on free and No on paid.
- Meta does not publish token prices for Muse Spark on its developer site.
Every price here was correct when this site last checked it, between 16 Sep and 22 Sep 2026. Prices change often and this site no longer updates them, so check the vendor’s own page before you rely on one: Google pricing · xAI pricing · Meta pricing.
Text models, per million tokens
| Model | Input | Output | Context |
|---|---|---|---|
| Gemini 3.8 Flash (to 31 Dec 2026) | $0.75 | $3.75 | — |
| Gemini 3.8 Flash (from 1 Jan 2027) | $1.50 | $7.50 | — |
| Gemini 3.1 Pro Preview (≤200k) | $2.00 | $12.00 | — |
| Grok 4.6 (<200k) | $2.00 | $6.00 | 500k |
| Grok 4.5 (<200k) | $2.00 | $6.00 | 500k |
| Grok 4.3 (<200k) | $1.25 | $2.50 | 1M |
Google, Gemini Developer API pricing, read in a browser 16 Sep 2026 · xAI, Pricing, read at source 16 Sep 2026.
The headline rate is not the number to plan on. Google’s cheapest rate has an expiry date on the page, and both vendors charge more once a prompt crosses a length threshold — xAI double, Google double on input.
The threshold that catches people
Both companies price long prompts higher, and both apply the higher rate to the whole request. xAI is explicit: “requests whose prompt reaches the listed token threshold are billed at the higher rate for all tokens in the request.”xAI, Pricing, read at source 16 Sep 2026: grok-4.6 is “$2.00” input below 200k prompt tokens and “$4.00” at or above it. Google’s Pro tier does the same at 200k, charging “$4.00, prompts > 200k tokens” against $2.00 below it.
So one long document in the context does not cost a little more. It re-prices everything in that call.
What the free tier costs
Google’s free tier is real and it is labelled. Every model table carries a row reading “Used to improve our products”, marked Yes on free and No on paid. The full split between the free console and the enterprise platform is on AI Studio vs Agent Platform.
xAI’s pricing page lists no free tier for the API at all — only prices.
That is a statement about the page, not about the company. A free option may exist elsewhere; it is not on the pricing page read here.
Why Meta is not in the table
Update, 22 Sep 2026: Meta now lists Muse Spark 1.3 at $1.25 per million input tokens and $4.25 per million output, covered on Meta’s developer stack; this table has not been rebuilt around it.Meta, Meta Model API, read at source 22 Sep 2026: “muse-spark-1.3 Not used to improve our products. 1M $1.25 $0.15 $4.25” (context window, then input, cached input and output per million tokens).
When this page was checked on 16 Sep 2026, the number was not published where it could be checked. Meta’s developer site advertises the Meta Model API as “Direct, self-serve and affordable access to Muse Spark” without naming a token price, and this site could not find a pricing page on it.Meta, Products & Solutions for AI Developers, read in a browser 16 Sep 2026; the site is script-rendered and refuses automated fetches.
Two Meta prices are published there: “Muse Image is now on Meta Model API at $0.01/image”, and Muse Voice Transcribe at “$0.18/hour”.Meta, AI developer site, read in a browser 16 Sep 2026: “Leading streaming transcription at $0.18/hour”.
Third-party sites quote per-token figures for Muse Spark. This page does not repeat them, because the rule here is that a price comes from the company charging it.
The costs that are not tokens
Both vendors bill separately for tools. xAI charges per thousand calls — web search and code execution at $5, file attachments at $10 — and since 21 September 2026 X Search has a different meter: “Starting September 21, 2026 at 12:00 PM PT, X Search is billed at $5 per 1k posts fetched and $10 per 1k user profiles fetched, replacing the current $5 per 1k calls.” The newer Grok 4.7 is priced like 4.6 (see Grok 4.7 vs 4.6).xAI, Pricing, read at source 16 Sep 2026: “Web Search” and “Code Execution” at “$5” per 1k calls, “File Attachments” at “$10”.
Google charges for grounding: “5,000 free search requests per month (shared across all Gemini 3.x models), then $14 per 1,000 requests.”Google, Gemini Developer API pricing, read in a browser 16 Sep 2026: the Gemini 3.8 Flash table’s “Grounding with Google Search” row.
Free or paid? These are API prices; for the Gemini and Grok consumer plans see free vs paid AI chat.
What this page could not verify
- Meta’s token prices. Not published on its developer site; secondary quotes are not used here.
- Whether the January increase holds. It is on the page today; pages change.
- Real cost per task. Price per token says nothing about how many tokens a model spends.
- Enterprise and volume pricing. Both vendors route that to sales.
Google, Gemini Developer API pricing, read in a browser on 16 September 2026 · xAI, Pricing, read at source 16 September 2026 · Meta, AI developer site, read in a browser 16 September 2026.
API prices are the most volatile figures on this site. Re-read before you budget.