Gemini 4 Carbon: API Pricing and Cursor Access
Gemini 4 Carbon has a stable paid API and a Cursor listing. Compare introductory and December rates, output limits, caching and separate Cursor billing.
By AI Pricing Guru Editorial Team
The AI Pricing Guru editorial team verifies prices against official provider pages and maintains the datasets, calculators and guides. See our methodology and contact us to report a correction.
Report summary — 2026-11-10
- Carbon is now documented as a stable paid Gemini API endpoint and listed among Cursor's third-party models. Google's announcement was October 28, not today.
- Introductory input, cached-input and output rates run through November 30, 2026; all three double on December 1.
- On current Cursor Premium-and-above plans, Carbon costs prepaid credit from the first request, outside the weekly pool. Free and Plus do not include third-party models.
- Teams and Enterprise eligible requests add the Cursor Token Rate. Carbon does not qualify for the strictly-below-threshold cache-price exemption.
Carbon introductory rates versus December standard pricing
New rates take effect December 1, 2026. Exact transition time/timezone not specified by Google. Prices are USD per 1M tokens.
| Model | Rate period | Input | Cached | Output | Input / output vs current |
|---|---|---|---|---|---|
| Gemini 4 Carbon | Current introductory rate | $2.00 | $0.1 | $10.00 | Comparison baseline |
| Gemini 4 Carbon | Standard rate from December 1, 2026 | $4.00 | $0.2 | $20.00 | 100% higher / 100% higher |
Built from the scheduled rate card in the canonical pricing API.
Cursor Teams billing and the additional token surcharge
| Plan | Recorded price | Tier | Published limits and access |
|---|---|---|---|
| The X Subscription Teams | From $30/seat/mo Price: Source verified: 2026-11-12 01:07:56 UTC. Official source | business | One workspace and bill with mixed seat tiers: Premium $30/seat/month, Super $100/seat/month, Ultra $200/seat/month. Each paid seat has the matching individual usage (Premium 4x, Super 15x, Ultra 50x Plus) in its own weekly Grok/Cursor pool; seats do not share pools. The team shares prepaid Additional Usage credits. Third-party Cursor models cost credits from the first request. Plus is not available for Teams. Annual team pricing is not published; no specific API model SKU is promised. On Teams and Enterprise, eligible third-party requests also incur the Cursor Token Rate of $0.25 per million input, output, and cached tokens on top of API rates, including BYOK. Cursor models, Cost mode, and the documented excluded models are exempt; individual plans do not pay this surcharge. Access and limits: Source verified: 2026-11-12 01:07:56 UTC. Official source |
Price and access verification are shown separately for each plan. Prices and plan notes are rendered at build time from the maintained subscription dataset.
Estimate Carbon and Argon token costs before Cursor fees and tools
Assumes 75% input tokens and 25% output tokens using current per-million rates.
Gemini 4 Carbon
Scheduled pricing effective December 1, 2026. Standard rate from December 1, 2026: $4 input / $0.2 cached input / $20 output per 1M tokens. Exact transition time/timezone not specified by the provider.
$40.00
- Input share
- $15.00
- Output share
- $25.00
Gemini 4 Argon
Scheduled pricing effective December 1, 2026. Standard rate from December 1, 2026: $4 input / $0.2 cached input / $20 output per 1M tokens. Exact transition time/timezone not specified by the provider.
$40.00
- Input share
- $15.00
- Output share
- $25.00
Carbon standard API token rates
| Model | Provider | Input / 1M | Cached / 1M | Output / 1M |
|---|---|---|---|---|
| Gemini 4 Carbon Price: Source verified: 2026-11-12 01:12:46 UTC. Official source | $2.00 | $0.1 | $10.00 |
Built from pricing.json at publish time.
Dated story text, scoped to the report date above:
What changed: current access, not a new launch
Google’s current model specification labels gemini-4-carbon stable, and its rate card lists paid API pricing with no Free tier. Cursor’s model catalog now lists Carbon as a third-party model.
The October 28 announcement described an initial restricted rollout to cyber defenders and trusted partners. That historical announcement is not today’s API eligibility rule, nor evidence of a November 10 launch. These current listings were checked November 10; they do not establish the exact day broader access began.
API pricing: budget beyond the introductory period
The tables above read Carbon’s current and scheduled rates directly from our canonical dataset. The introductory period ends November 30, 2026, with standard rates starting December 1. Google gives no precise transition timezone.
Output billing includes thinking tokens. Internal tool system instructions and function declarations can also count toward billed tokens when grounding is used; tool charges may be additional. The inline calculator compares Carbon and Argon using uncached input/output only, excluding Cursor fees, tools and retries. Reprice workloads that will continue into December rather than projecting the introductory discount indefinitely.
Caching is implicit only: explicit cache creation is unsupported. Do not assume every repeated input earns the cache-read rate; budget using actual cache-hit telemetry.
Cursor: available does not mean included
On current Premium and above, third-party usage is charged at API list price from prepaid credit from the first request. It does not first consume the weekly Grok/Cursor-model pool. Free and Plus exclude third-party models. Legacy Cursor plans retain different billing rules; check your account’s actual plan before applying the current-plan explanation.
For Teams and Enterprise, the additional Cursor Token Rate applies to eligible input, output and cached tokens, including eligible BYOK requests. Its maintained amount is recorded in the Teams snapshot above; individual plans do not pay this fee.
The exemption requires a cache-read price strictly below Cursor’s threshold across every pricing tier. Carbon’s introductory cache rate equals that threshold, rather than falling below it, and its scheduled rate is higher: Carbon is not exempt on that basis. Cursor models and Cost mode have separate exemptions. See the official surcharge rules.
Paid API availability and a Cursor listing do not prove inclusion in Google AI Pro, Ultra, or any other consumer Gemini subscription.
Context limits and practical next steps
Google specifies 1,048,576 tokens of unified input/output context. Default synchronous output is 65,536 tokens; long decode can extend output up to the shared context ceiling minus input tokens, not a separate million-token output allowance on top of a full prompt.
- API developers: use the stable endpoint, check project quotas, and measure thinking tokens, cache hits and successful-task cost before increasing traffic.
- Cursor individuals: confirm Premium-or-higher eligibility and set an additional-usage cap; a larger weekly pool does not make Carbon included usage.
- Team admins: include the token surcharge in budgets and audit routing, cached-token volumes and shared prepaid-credit controls.
See Google model pricing and OpenAI pricing for alternatives, the token calculator for custom volumes, and our API cost guide for workload budgeting.