Inference
EN
Google

Gemini 4 Carbon

Gemini 4 Carbon is a reasoning model from Google in the Gemini 4 family. 28 benchmarks count toward its index, in 5 categories.

  • Closed weights
  • Reasoning
  • Input: audio · file · image · text · video
  • Released 2026-10
RankedEnough results for an index.
IndexAll categories, out of 100.82.7±5.1Median 51.0
CoverageIndex weight that has results.80%
SpeedOutput tokens per second.31/sMedian 67/s
First tokenSeconds to the first output token.8.1 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$2.00batch $1.00Median $0.60
Output / 1MUS dollars per 1M output tokens.$10.00batch $5.00 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1538±10 #1Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
google/gemini-4-carbon
Max outputMaximum output tokens per request.
944K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.10
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
3 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Google AI StudioMaker$2.00$10.001.05M
Google Vertex AIMaker$2.00$10.001.05M
OpenRouterRouter$2.00$10.001.05M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.10 counted83.21 / 218
CodingCode writing and repair.4 counted80.54 / 209
ReasoningLogic problems and puzzles.8 counted83.92 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.3 counted79.52 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.None counted——
MathMath problems.3 counted77.25 / 168

Results

28 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
GPQA diamond · Epoch AIKnowledge96.5% ±1.21 / 217
OTIS Mock AIME 2024-2025Math100.0%1 / 200
SimpleQA VerifiedKnowledge77.7% ±1.32 / 91
Chess PuzzlesReasoning78.0% ±4.22 / 90
Agent Arena steerabilityAgentic19.12 / 54
ARC-AGI-2 (semi-private)Reasoning92.1%5 / 131
ARC-AGI-1 (semi-private)Reasoning97.5%5 / 126
Artificial Analysis AutomationBenchAgentic77.8%1 / 23
ExploitGymAgentic53.5%1 / 23
cursorBench40Coding60.4%1 / 21
Artificial Analysis Omniscience AccuracyKnowledge58.3%9 / 164
FrontierMath-Tier-4-v2-PrivateMath95.1% ±3.44 / 72
Mystery Game PuzzlesReasoning73.0% ±4.55 / 80
EBR-benchReasoning87.6% ±6.82 / 31
Terminal-Bench 4.0Agentic63.1% ±2.24 / 61
FrontierMath-Tiers-1-3-v2-PrivateMath90.5% ±1.76 / 90
Vibe Code Bench v1.1Coding89.1% ±2.39 / 123
Agent Arena task outcomeAgentic18.44 / 54
ARC-AGI-3 (semi-private)Reasoning63.9%2 / 26
CWE-bench v1Agentic71.0%1 / 12
Artificial Analysis Terminal-Bench v4.0Agentic65.2%2 / 22
Multimodal Knowledge Agent BenchmarkAgentic54.3%2 / 19
AutomationBenchAgentic53.9%5 / 32
Furniture AssemblyReasoning77.5% ±5.46 / 37
FrontierCode 2.0Coding40.0%5 / 27
FrontierSWE v2Coding61.2%5 / 22
Artificial Analysis Long Context ReasoningReasoning79.3%52 / 217
Artificial Analysis GDP.pdfAgentic28.4%7 / 22
OSWorld 2.1 offline subset (82 tasks)Agentic81.5%5 / 10
Legal Research BenchLegal55.8% ±3.51 / 90
Tax Agent BenchFinance51.8% ±3.61 / 82
ARC-AGI-2 (public eval)Reasoning97.9%3 / 122
ARC-AGI-1 (public eval)Reasoning98.5%6 / 116
FrontierMath-ErdosMath2.9% ±2.12 / 15
HLAB-Vals-1.1Legal17.5% ±3.08 / 35
GraphWalks BFS 256K–1MReasoning96.2%—
Chartography without toolsMultimodal72.5%—
AA-Briefcase v1.1, rubric pass rateAgentic61.8%—
Blueprint-Bench 2Multimodal58.6%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseEpoch AICC BYARC Prize FoundationTerms restrict reuseVals AINo licence statedLMArenaCC BY 4.0models.devMIT

History

5 releases
Index82.3 → 82.7
29 Oct 202612 Nov 2026
Output / 1M$10.00 → $10.00
29 Oct 202612 Nov 2026

Questions

How much does Gemini 4 Carbon cost?
$2.00 per 1M input tokens, $10.00 per 1M output tokens.
How good is Gemini 4 Carbon?
Its index is 82.7 out of 100. That is place 2 of 283 ranked models.
What is Gemini 4 Carbon best at?
Agentic: 83.2 out of 100, place 1 of 218.
How fast is Gemini 4 Carbon?
31 output tokens per second, 8.1 s to the first token.
What is the context window of Gemini 4 Carbon?
1.05M tokens.
Is Gemini 4 Carbon open source?
No. Its weights are not published.
When was Gemini 4 Carbon released?
October 2026.

Badge

Gemini 4 Carbon on Inference360: place 2 of 283, index 82.7
HTML
<a href="https://inference360.ai/models/gemini-4-carbon"><img src="https://inference360.ai/badges/gemini-4-carbon.svg" alt="Gemini 4 Carbon on Inference360: place 2 of 283, index 82.7" height="20"></a>
Markdown
[![Gemini 4 Carbon on Inference360: place 2 of 283, index 82.7](https://inference360.ai/badges/gemini-4-carbon.svg)](https://inference360.ai/models/gemini-4-carbon)

Compare with

Claude Fable 5.285.6Claude Opus 5.578.6Gemini 4 Argon78.5GPT-6 Astra78.0GPT-6.1 Sol78.0

More from Google

All models
Gemini Omni 1.1 FlashUnrankedGemini Omni 1.2 FlashUnrankedGemini Decision 1UnrankedNano Banana 2.1UnrankedEmbeddingGemma 2Unranked