Vascend
HorizonVascend IDEVascend CLICommunityBenchmarkContact
Sign inJoin Discord
HomeHorizonIDECLISign in
ChangelogIntelligenceCost per taskCoding Agent IndexCyber IndexDecision IndexBenchmarksSpeedLatencyPriceLeaderboard
Benchmark

Gemma 4 12B (Reasoning)

  • Google
  • Release · June 3, 2026
  • Open weights
  • Reasoning

Intelligence

Intelligence Index

#127 of 266
  1. 15.215Qwen3.6 35B A3B (Non-reasoning)
  2. 14.715Gemma 4 31B (Reasoning)
  3. 14.314ERNIE 5.0 Thinking Preview
  4. 14.214Nova 2.0 Pro Preview (Medium)
  5. 14.214Mistral Medium 3.5
  6. 14.214Gemma 4 12B (Reasoning)
  7. 13.914Gemma 4 31B (Non-reasoning)
  8. 13.614Nova 2.0 Omni (Medium)
  9. 13.413Nova 2.0 Lite (High)
  10. 13.413Apriel-v1.6-15B-Thinker
Data from Artificial Analysis

Evaluations

  • AA-Omniscience Accuracy15.6%#165
  • Humanity's Last Exam15.7%#116
  • AA-LCR63.7%#122
  • CritPt0%#143
  • MMMU-Pro69.7%#64
  • GPQA Diamond75.3%#91
  • IFBench73.5%#13
  • τ²-Bench Telecom36.3%#79
  • Terminal-Bench Hard18.2%#37
Gemma 4 12B (Reasoning)Median
Data from Artificial Analysis

Speed and price

Speed

tokens per second · #76 of 162
  1. 118118Gemma 4 12B (Non-reasoning)
  2. 116116GPT-6 Luna (Low)
  3. 116116GPT-5.6 Terra (Low)
  4. 115115Nova 2.0 Pro Preview (Non-reasoning)
  5. 114114GPT-5.6 Terra (Medium)
  6. 114114Gemma 4 12B (Reasoning)
  7. 114114Agnes 2.5 Pro Beta
  8. 113113MiniMax-M3.1-Flash
  9. 112112Solar Mini 4
  10. 109109Mistral Large 4 Preview
Data from Artificial Analysis

Latency

seconds to first token · #87 of 162
  1. 2.12.1Qwen3.8-Flash-Next
  2. 2.22.2Qwen3.5 397B A17B (Non-reasoning)
  3. 2.22.2Qwen3.5 397B A17B (Reasoning)
  4. 2.22.2K2 Horizon 375B A23B
  5. 2.22.2GPT-5.6 Terra (High)
  6. 2.22.2Gemma 4 12B (Reasoning)
  7. 2.32.3Gemma 4 12B (Non-reasoning)
  8. 2.32.3Qwen3.5 122B A10B (Reasoning)
  9. 2.32.3Hermes 4 - Llama-3.1 405B (Non-reasoning)
  10. 2.32.3Qwen3.7 Plus
Data from Artificial Analysis

Price

$ per 1M tokens (3:1) · #27 of 169
  1. $0.130.13Apertus 8B Instruct
  2. $0.130.13gpt-oss-20b (Low)
  3. $0.130.13Phi-4 Mini Instruct
  4. $0.140.14Phi-4 Multimodal Instruct
  5. $0.150.15Agnes 2.5 Pro Beta
  6. $0.150.15Gemma 4 12B (Reasoning)
  7. $0.150.15Gemma 4 12B (Non-reasoning)
  8. $0.150.15Ministral 3 8B
  9. $0.150.15Qwen3.5 9B (Reasoning)
  10. $0.160.16Gemma 4 26B A4B (Reasoning)
Data from Artificial Analysis

Nearby models

  • Qwen3.6 35B A3B (Non-reasoning)Alibaba15.2
  • Gemma 4 31B (Reasoning)Google14.7
  • ERNIE 5.0 Thinking PreviewBaidu14.3
  • Nova 2.0 Pro Preview (Medium)Amazon14.2
  • Mistral Medium 3.5Mistral14.2
  • Gemma 4 31B (Non-reasoning)Google13.9
  • Nova 2.0 Omni (Medium)Amazon13.6
  • Nova 2.0 Lite (High)Amazon13.4
Data from Artificial Analysis

Same lab

  • Gemini 4 Argon (High)Google52.6
  • Gemini 3.8 Flash (High)Google40.9
  • Gemini 3.8 Flash (Medium)Google39.8
  • Gemini 3.8 Flash (Low)Google33.5
  • Gemini 3.1 Pro PreviewGoogle29.7
  • Gemini 3.5 Flash-LiteGoogle22.2
  • Gemma 4 26B A4B (Reasoning)Google16.7
  • Gemma 4 26B A4B (Non-reasoning)Google13.1
Data from Artificial Analysis
HorizonVascend IDEOlympusTalosCoursesCommunityBenchmark
DiscordGitHubXLet’s talkLegalPrivacyCookiesData useSecurity
© 2026 Vascend Technologies · VAT IT04911550616
ascend