TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Google opened Gemini 4 Carbon to paying developers, businesses and Google AI Ultra subscribers on November 10. On the new MKA benchmark, Carbon scored 54.3%, behind Claude Fable 5.2 at 56.6%; Google DeepMind reported a lower cost per task for Carbon, though the benchmark was co-developed by Google and its tasks are private.
Google opened Gemini 4 Carbon to paying developers, businesses and Google AI Ultra subscribers on November 10, the same day Google DeepMind and Artificial Analysis introduced MKA, a benchmark for AI agents handling materials such as blueprints, scans and charts. Carbon scored 54.3%, second to Anthropic’s Claude Fable 5.2 at 56.6%; Google DeepMind’s report put Carbon’s cost at $5.63 per task, compared with $35.28 for Fable 5.2.
Carbon is Google’s top-tier model, built on Gemini 4 Argon. Google describes it as intended for demanding work such as complex coding and debugging and open-ended scientific research. Its model card says it uses more thinking tokens than Argon, which can mean more extensive reasoning and verification but also higher token use. Google says Argon remains the preferred choice for most software engineering and knowledge work.
Access depends on the product. Carbon is available through the paid Gemini API, Google AI Studio and Gemini for business. In the Gemini app and AI Mode in Search, it is limited to Google AI Ultra subscribers. Google lists pricing at $2 per million input tokens and $10 per million output tokens through November 30, rising to $4 and $20 from December 1. Google’s model card reports that Carbon led on 15 of 17 benchmarks it selected, but that comparison does not include Claude Fable 5.2.
MKA comprises 200 expert-built tasks across 10 sectors. The published scores put Fable 5.2 at 56.6%, Carbon at 54.3% and Claude Opus 5.5 at 53.0%. The source report says these top-three results are within the margin of error. Google DeepMind reported MKA costs of $5.63 per Carbon task and $35.28 per Fable 5.2 task; these are benchmark-specific figures, not a general estimate of what every user will pay.
Cost and Scores on MKA
The MKA results put attention on two questions for businesses considering AI agents: how well a model handles information that is not plain text, and what it costs to complete a task. Carbon finished 2.3 percentage points behind Fable 5.2 on the reported scores, while Google DeepMind’s cost figures show a substantial difference for the benchmark’s tested tasks. That may make Carbon worth evaluating for workloads where the quality gap and task cost both matter.
The figures do not establish which model is best for a particular company. MKA covers a defined set of 200 tasks, and results can vary across real-world documents, prompts and workflows. The benchmark’s reported costs also should not be treated as a direct forecast for every deployment. Organizations would need to assess their own tasks, accuracy needs, latency and total usage before drawing purchasing conclusions.
Products Worth Considering
DJI Mini 4K Drone with 4K UHD Camera
Bose QuietComfort Headphones (2nd Gen), Wireless Over-Ear Noise-Cancelling Headphones, Black
Sonos Arc Ultra Soundbar with Dolby Atmos, Black
As an affiliate, we earn on qualifying purchases.
From Limited Access to Paid Use
Carbon was initially made available to a small group of cyber defenders through Google’s Fairwind Program, with Artificial Analysis listing October 28 as its release date. Its stronger security features remain limited to verified defenders, according to the source material. The November 10 announcement expanded access to paying users and included publication of Google’s model card.
Carbon’s independent standing has shifted as new models and comparisons appeared. On Artificial Analysis’s Intelligence Index v5.0.1, Carbon at its high setting scored 60. Claude Fable 5.2 scored 65 at its maximum setting and 62 at high. The index uses nine tests, according to Artificial Analysis. Google’s model card instead presents results from tests Google selected and says those results are as of October 2026; its table predates Fable 5.2’s November 9 release.
““tasks that test the limits of a model’s intelligence””
— Google DeepMind, Gemini 4 Carbon model card
Limits of the MKA Comparison
Google co-developed MKA, and the benchmark’s tasks are private, according to the source material. That limits outsiders’ ability to inspect the task set or independently reproduce the results. The published scores are therefore evidence about performance on this benchmark, but readers cannot use the available materials to check how representative its tasks are of broader work.
The reported top-three scores are within the margin of error, so the ordering should be read cautiously. The source does not give the margin’s size or further detail on how it was calculated. Google also lists possible Carbon limitations, including hallucinations, occasional slowness or timeouts, and greater token use at higher thinking settings. How often those issues arise in practical use is not established here.
Pricing Changes in December
Google’s listed introductory API pricing runs through November 30, 2026. From December 1, it is scheduled to double to $4 per million input tokens and $20 per million output tokens. Developers and businesses evaluating Carbon can compare its performance and usage costs on their own workloads before that change takes effect.
Further evidence will depend on use beyond MKA, including how Carbon performs on tasks customers choose and how its reliability and speed hold up in deployment. The published information does not say when MKA’s tasks may become available for outside review or when another benchmark update will appear.
Key Questions
Who can use Gemini 4 Carbon?
It is available through the paid Gemini API, Google AI Studio and Gemini for business. In the Gemini app and AI Mode in Search, access is limited to Google AI Ultra subscribers.
How did Carbon score on the MKA benchmark?
Carbon scored 54.3%, behind Claude Fable 5.2 at 56.6% and ahead of Claude Opus 5.5 at 53.0%. The source report says the top three scores are within the margin of error.
What does Google say Carbon costs per MKA task?
Google DeepMind’s report lists $5.63 per Carbon task and $35.28 per Fable 5.2 task. These are reported benchmark costs; they do not establish what a customer will pay for a different workload.
When does Carbon’s listed API pricing change?
The listed rate is $2 per million input tokens and $10 per million output tokens through November 30, 2026. Google’s stated rates from December 1 are $4 per million input tokens and $20 per million output tokens.
Source: Thorsten Meyer AI
