Gemini 4 Carbon
by Google
Today's price · per 1M tokens
Input
$2.00
per 1M tokens
Output
$10.00
per 1M tokens
Blended
$4.00
blended $/1M — 3:1 weighted input:output
Cached input
$0.100
5% of input — prompt caching
How this price is scoped: Google's Gemini API pricing page prints a dated, two-phase schedule for this model on its Standard tier: $2.00 per 1M input tokens through November 30, 2026, rising to $4.00 on December 1, 2026, and $10.00 per 1M output tokens (thinking tokens included), rising to $20.00. Context caching follows the same dates ($0.10, then $0.20; implicit caching only). We publish the rate you would pay today and re-verify it before the changeover. Batch and Flex requests cost half the Standard rate. Google also sells a Priority tier at $3.60 input and $18.00 output through November 30 ($7.20 and $36.00 from December 1). We publish the Standard rate.
Price receipt
We read Google's own pricing page on and found Gemini 4 Carbon listed with a price on the same card — that read is where the number above comes from, recorded as a new model. We keep the page text we read; its content hash is 117ebf5ad3.
Overview
Google's most capable reasoning model, generally available since 2026-11-10, built for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents and long-decode generation. Currently $2/1M input and $10/1M output on Google's introductory schedule, rising to $4/$20 on 2026-12-01. 1M-token context with native image, audio and video input.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterWe hold no per-benchmark accuracy scores for this model yet, so it has no accuracy average. That is a gap in our coverage, not a sign the model is untested — independent evaluators often publish a composite index for a new model long before releasing its per-benchmark numbers. The percentile above is its standing across the independent composites below.
- AA Intelligence Index: 60 (high effort) (Artificial Analysis composite across reasoning, knowledge and coding evals)
- AA-Omniscience Index: 53.5 (high effort) (−100–100)
- GDPval-AA v2: 1732.7 ELO (Real-world work tasks, human baseline = 1000)
- Arena (arena.ai): 1526.2 ELO (Human preference)
- Arena WebDev: 1768.2 ELO (Human preference on building web apps)
- Arena Agent: 0.17 (Score from real agent sessions (IPS); higher is better)
- Arena Agent cost per task: $1.46 (Median measured cost of one real Agent Arena task)
- Epoch Capabilities Index: 169.8 (CI 166.5–173.5) (Epoch AI's index fitted across many benchmarks)
- FrontierMath Tiers 1–3: 90.5% (high effort) (Research-level maths (funded by OpenAI, run by Epoch); best run across the effort settings Epoch tested)
- FrontierMath Tier 4: 95.1% (high effort) (Hardest research maths; best run across the effort settings Epoch tested)
Source: Artificial Analysis (artificialanalysis.ai)
- AA-LCR (long-context reasoning): 79.3% (high effort)
- Terminal-Bench 4.0: 65.2% (high effort)
- Cost to run the AA Intelligence Index: $5.49 per task, 29 min per task (high effort)
- Measured by AA: 70.6 tokens/sec output, 66.65s to first token (high effort)
- Ranks #2 of 189 comparably-measured models by percentile score, across 4 independent measurements
- Ranks #119 of 189 comparably-tested models by normalized performance per dollar
70.6 tokens/sec output 66.65s latency to first token (TTFT)
Throughput: Artificial Analysis, measured 2026-11-14 · Source: Artificial Analysis (artificialanalysis.ai)
Source: Artificial Analysis · updated · See full rankings →
Specifications
- Provider
- Context window
- 1M tokens
- Modality
- text, image, audio, video
- Parameters
- Proprietary
- Open source
- No — proprietary
- Released
- Status
- Current
- Last updated
- Tags
Availability verified: — listed on Google's own page