Gemini 4 Argon

  • Closed weights
  • Reasoning
  • Input: audio · file · image · text · video
  • Released 2026-09
RankedEnough results for an index.
IndexAll categories, out of 100.76.8±3.2Median 50.7
CoverageIndex weight that has results.90%
SpeedOutput tokens per second.30/sMedian 76/s
First tokenSeconds to the first output token.4.2 sMedian 1.1 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 $2.50
ContextMaximum tokens in one request.1.05MMedian 262K
EloLMArena rating and rank.1528±5 #2Median 1425

Details

From OpenRouter
API IDThe model ID on OpenRouter.
google/gemini-4-argon
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
Weights
—
RetiresDate OpenRouter removes the model.
—

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.18 counted75.48 / 221
CodingCode writing and repair.15 counted75.212 / 214
ReasoningLogic problems and puzzles.10 counted77.512 / 225
MultimodalTasks with images and text.2 counted79.52 / 88
KnowledgeFacts and expert knowledge.4 counted75.210 / 266
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.None counted——
MathMath problems.4 counted74.414 / 171

Results

53 counted
BenchmarkResultThe score from the source.PlacePlace among results on this benchmark.
OTIS Mock AIME 2024-2025Math100.0%1 / 203
ARC-AGI-1 (semi-private)Reasoning98.5%1 / 132
LiveBench ReasoningReasoning95.1%1 / 80
LiveBench Data AnalysisReasoning85.2%1 / 80
GPQA diamond · Epoch AIKnowledge96.4% ±1.24 / 220
IOICoding100.0%1 / 46
Code MigrationCoding68.2% ±4.32 / 77
Vibe Code Bench v1.1Coding91.9% ±1.94 / 130
SimpleQA VerifiedKnowledge76.0% ±1.44 / 94
Artificial Analysis SciCodeCoding61.8%5 / 113
DeepSWEAgentic77.9%3 / 64
ARC-AGI-2 (semi-private)Reasoning91.3%9 / 137
LiveBench Browser UseAgentic92.0%4 / 49
Critical Physics TasksReasoning27.1%17 / 207
LiveBench Long AgenticAgentic69.9%5 / 51
GDPval-AA normalizedAgentic56.2%14 / 132
ProofBench v1.1Math99.0% ±1.06 / 55
Mystery Game PuzzlesReasoning64.0% ±4.811 / 83
PostTrainBench v1.1Coding45.3%2 / 15
Chess PuzzlesReasoning59.0% ±4.913 / 93
Agent Arena task outcomeAgentic12.69 / 59
Terminal-Bench 4.0Agentic57.6% ±2.311 / 68
OSWorld 2.0Agentic69.2%4 / 24
CWE-bench v1Agentic68.0%2 / 12
FrontierMath-Tiers-1-3-v2-PrivateMath86.0% ±2.116 / 93
ExploitGymAgentic39.5%4 / 23
FrontierMath-Tier-4-v2-PrivateMath78.0% ±6.515 / 75
Agent Arena steerabilityAgentic12.212 / 59
PostTrainBench v1.2Coding40.4%5 / 23
LiveBench ResearchAgentic83.1%12 / 52
ARC-AGI-3 (semi-private)Reasoning46.3%7 / 30
EBR-benchReasoning64.8% ±4.98 / 33
Furniture AssemblyReasoning72.5% ±5.610 / 40
Chartography without toolsMultimodal71.6%3 / 12
Terminal-Bench 4.1.0Agentic55.5% ±1.513 / 52
cursorBench40Coding52.9%6 / 21
LiveBench Coding 2.0Coding89.7%15 / 52
FrontierSWE v2Coding55.0%8 / 25
Vibe Code Bench 1-100Coding21.2% ±3.89 / 26
MirrorCodeCoding58.9% ±11.714 / 40
AutomationBenchAgentic51.3%13 / 33
Agents' Last ExamAgentic39.5%9 / 21
Bug Hunt BenchCoding40.7%13 / 30
CWE-bench v2Agentic65.0%9 / 20
cursorBench41Coding56.5%6 / 12
FrontierCode 2.0Coding27.6%18 / 32
Terminal-Bench 2.1 · Vals AIAgentic57.7% ±2.543 / 75
LiveBench DesignAgentic73.7%32 / 52
Agent Arena command recoveryAgentic−21.448 / 50
ARC-AGI-1 (public eval)Reasoning99.5%1 / 116
Finance Agent (v2)Finance65.4% ±0.31 / 79
LegalBenchLegal88.3% ±0.43 / 149
HLAB-Vals-1.1Legal27.1% ±3.71 / 42
MedCodeHealthcare58.8% ±2.13 / 108
EMBFinance75.2% ±2.54 / 75
CyberBench v1.1Other77.9% ±5.33 / 49
Legal Research BenchLegal54.8% ±3.56 / 98
ARC-AGI-2 (public eval)Reasoning97.1%8 / 127
Tax Agent BenchFinance49.2% ±3.66 / 89
SRE BenchOther44.3% ±3.13 / 37
ProgramBenchCoding2.5% ±1.18 / 66
SAGEEducation89.5% ±1.64 / 30
Public Benefits Bench v1.1Public benefits69.8%7 / 50
MedScribeHealthcare87.4% ±2.018 / 111
Terminal-Bench-ScienceScience44.3% ±6.09 / 48
MysteryMechanismScience45.5% ±3.48 / 34
BioMysteryBenchScience76.3% ±0.410 / 34
Time Horizon Index: KSPOther52.5% ±8.94 / 13
FrontierMath-ErdosMath1.5% ±1.57 / 16
CUA-benchOther4.8%8 / 9
Graphwalks BFS 0K-128KReasoning99.7%—
LVBenchMultimodal91.7%—
GraphWalks BFS 256K–1MReasoning84.2%—
Blueprint-Bench 2Multimodal54.4%—
RoboDojo SimAgentic34.2%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedEpoch AICC BYARC Prize FoundationTerms restrict reuseLiveBenchNo licence statedTerminal-BenchNo licence statedLMArenaCC BY 4.0models.devMITCheaper InferenceNo licence stated

History

13 releases
Index81.0 → 76.8
23 Oct 202621 Nov 2026
Output / 1M$10.00 → $10.00
23 Oct 202621 Nov 2026

Questions

How much does Gemini 4 Argon cost?
$2.00 per 1M input tokens, $10.00 per 1M output tokens.
How good is Gemini 4 Argon?
Its index is 76.8 out of 100. That is place 6 of 282 ranked models.
What is Gemini 4 Argon best at?
Multimodal: 79.5 out of 100, place 2 of 88.
How fast is Gemini 4 Argon?
30 output tokens per second, 4.2 s to the first token.
What is the context window of Gemini 4 Argon?
1.05M tokens.
Is Gemini 4 Argon open source?
No. Its weights are not published.
When was Gemini 4 Argon released?
September 2026.

Badge

Gemini 4 Argon on Inference360: place 6 of 282, index 76.8

Compare with

GPT-6 Astra76.5Claude Opus 5.577.1Gemini 4 Hydrogen76.2GPT-6.1 Sol76.2GPT-6.2 Sol76.0

Lower price

GPT-6.1 Terra75.1 · $5.00/1MGemini 4 Hydrogen76.2 · $7.50/1M

More from Google

All models
EmbeddingGemma 2UnrankedNano Banana 2.1UnrankedGemini 3.8 LiveUnrankedGemini 3.8 Live Extended ThinkingUnrankedGemini 4 Carbon80.5