Vascend
Sign inJoin Discord
Benchmark

Gemini 4 Carbon (High)

  • Google
  • Release · October 28, 2026
  • Proprietary
  • Reasoning

Intelligence

Intelligence Index

#1 of 77
  1. 59.459Gemini 4 Carbon (High)
  2. 58.458Claude Opus 5.5 (Max)
  3. 5757Claude Opus 5.5 (Xhigh)
  4. 56.957GPT-6 Astra (Max)
  5. 55.355GPT-6 Astra (Xhigh)
  6. 54.855GPT-6 Astra (High)
  7. 54.555GPT-6.1 Sol (Max)
  8. 54.254Claude Sonnet 5.5 (Max)
  9. 52.352GPT-6.1 Sol (Xhigh)
  10. 51.752Claude Opus 5.5 (High)
Data from Artificial Analysis

Evaluations

  • AA-Briefcase v1.11,610#15
  • AA-Omniscience v1.1 Accuracy58.3%#21
  • AA-Omniscience v1.1 Non-Hallucination88.6%#2
  • AutomationBench-AA77.8%#1
  • Terminal-Bench-Science 0.164.8%#1
  • FrontierCode 2.040.6%#7
  • HLE-Diamond69.1%#1
  • GDP.pdf28.4%#15
  • Terminal-Bench 4.065.2%#2
  • CritPt v2 (Hard Subset)63%#10
  • GDPval-AA v2.11,733#11
  • FrontierCode 1.148.5%#7
  • AA-LCR v1.179.3%#63
  • SciCode64.2%#3
  • Humanity's Last Exam60.4%#2
  • CritPt36%#1
Gemini 4 Carbon (High)Median
Data from Artificial Analysis

Cost per task

Cost per task

$ per Intelligence Index task · #21 of 26
  1. $5.015.01GLM-5.3 (Max)
  2. $5.545.54Mistral Large 4 Preview
  3. $6.246.24GPT-6 Astra (Max)
  4. $6.956.95Muse Wonder (Max)
  5. $7.517.51Grok 4.7 (Xhigh)
  6. $7.767.76Gemini 4 Carbon (High)
  7. $7.777.77Gemini 3.8 Flash (High)
  8. $10.02$10Claude Sonnet 5.5 (Max)
  9. $10.02$10Qwen3.8 Max (0902)
  10. $10.04$10Claude Opus 5.5 (Max)
Data from Artificial Analysis

Time per task

seconds per task · #21 of 26
  1. 1,0331,033MiniMax-M3.1-Flash
  2. 1,0611,061GLM-5.3 (Max)
  3. 1,1471,147Mistral Large 4 Preview
  4. 1,1471,147Claude Opus 5.5 (Max)
  5. 1,1781,178Claude Fable 5.1 (Max)
  6. 1,2481,248Gemini 4 Carbon (High)
  7. 1,3261,326Claude Sonnet 5.5 (Max)
  8. 1,4361,436Kimi K3.1 Pro
  9. 1,6751,675MiMo-V2.6-Pro
  10. 2,0982,098Qwen3.8 27B (Xhigh)
Data from Artificial Analysis

Tokens per task

output tokens per task · #20 of 28
  1. 109.8K110KQwen3.8 27B (Xhigh)
  2. 113.5K114KStep 5 Preview
  3. 117.9K118KClaude Fable 5.1 (Max)
  4. 119K119KGLM-5.3 (Max)
  5. 121.1K121KMiMo-V2.6-Pro
  6. 122.8K123KGemini 4 Carbon (High)
  7. 135.3K135KMistral Large 4 Preview
  8. 138.2K138KMiniMax-M3.1-Flash
  9. 139K139KDeepSeek V4.1 Flash 1017 (Max)
  10. 170.8K171KQwen3.8 Max (0902)
Data from Artificial Analysis

Speed and price

Speed

tokens per second · #84 of 161
  1. 107107MiniMax-M3
  2. 106106Grok 4.7 (High)
  3. 105105Magistral Small 1.2
  4. 102102Qwen3 Omni 30B A3B (Reasoning)
  5. 101101Grok 4.7 (Xhigh)
  6. 100100Gemini 4 Carbon (High)
  7. 9898Claude Opus 5.5 (Max)
  8. 9797Gemma 4 26B A4B (Reasoning)
  9. 9797Qwen3 Omni 30B A3B Instruct
  10. 9292Ministral 3 8B
Data from Artificial Analysis

Latency

seconds to first token · #151 of 161
  1. 4545Gemini 4 Argon (High)
  2. 53.654Claude Haiku 5.5 (Xhigh)
  3. 56.657GPT-5.3 Codex (Xhigh)
  4. 73.974GPT-6 Luna (Max)
  5. 8383Claude Fable 5.1 (Xhigh)
  6. 97.598Gemini 4 Carbon (High)
  7. 101.9102GPT-6.1 Sol (Xhigh)
  8. 128.9129Claude Fable 5.1 (Max)
  9. 129.8130GPT-6 Astra (Xhigh)
  10. 131131Claude Opus 5.5 (Xhigh)
Data from Artificial Analysis

Price

$ per 1M tokens (3:1) · #151 of 169
  1. $44Claude Sonnet 5.5 (Medium)
  2. $44Claude Sonnet 5.5 (Low)
  3. $4.384.38Command A
  4. $4.54.5Gemini 3.1 Pro Preview
  5. $4.814.81GPT-5.3 Codex (Xhigh)
  6. $88Gemini 4 Carbon (High)
  7. $88Claude Opus 5.5 (Max)
  8. $88Claude Opus 5.5 (Xhigh)
  9. $88Claude Opus 5.5 (High)
  10. $88Gemini 4 Argon (High)
Data from Artificial Analysis

Nearby models

Data from Artificial Analysis

Same lab

Data from Artificial Analysis