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

Qwen3.8 27B (Xhigh)

  • Alibaba
  • Release · August 14, 2026
  • Open weights
  • Reasoning

Intelligence

Intelligence Index

#63 of 266
  1. 34.635GPT-6 Luna (Xhigh)
  2. 34.535Claude Haiku 5.5 (Medium)
  3. 34.334GLM-5.3 (Low)
  4. 34.234GPT-5.6 Terra (High)
  5. 33.934JT-4.1 Flash 236B A21B (Reasoning)
  6. 33.734Qwen3.8 27B (Xhigh)
  7. 33.634Motif 3
  8. 33.534Gemini 3.8 Flash (Low)
  9. 32.933GPT-6 Luna (High)
  10. 32.533GPT-5.3 Codex (Xhigh)
Data from Artificial Analysis

Evaluations

  • AA-Briefcase1,394#19
  • AA-Omniscience Accuracy15.6%#165
  • AA-Omniscience Non-Hallucination69.7%#10
  • AutomationBench-AA48.2%#23
  • GDPval-AA1,419#22
  • Humanity's Last Exam33.9%#71
  • GDP.pdf16.4%#20
  • Terminal-Bench 4.05.6%#63
  • AA-LCR82%#31
  • SciCode46.6%#82
  • CritPt5.4%#81
  • Terminal-Bench-Science0.5%#23
  • FrontierCode34.2%#21
  • EnterpriseOps-Gym-AA44.2%#3
  • MLCR-AA22.8%#8
  • MMMU-Pro76.3%#44
  • PersonalityBench: Expressiveness13.7#26
  • PersonalityBench: Organization56.8#16
  • PersonalityBench: Transparency31.5#22
  • PersonalityBench: Slop84.8#3
  • PersonalityBench: Informativeness28.9#20
  • PersonalityBench: Warmth23.8#23
  • PersonalityBench: Conviction31.5#21
  • PersonalityBench: Imagination34#22
  • GPQA Diamond90.5%#24
  • τ²-Bench Banking48%#3
Qwen3.8 27B (Xhigh)Median
Data from Artificial Analysis

Cost per task

Cost per task

$ per Intelligence Index task · #11 of 24
  1. $0.230.23Beam 1.0
  2. $0.260.26DeepSeek V4.1 Flash 1017 (Max)
  3. $0.280.28GLM-5.3-Flash
  4. $0.720.72GPT-6.1 Sol (Max)
  5. $0.840.84Muse Spark Open (Xhigh)
  6. $1.011.01Qwen3.8 27B (Xhigh)
  7. $1.031.03Step 5 Preview
  8. $1.131.13Mistral Large 4 Preview
  9. $1.241.24Gemini 3.8 Flash (High)
  10. $1.991.99Gemini 4 Argon (High)
Data from Artificial Analysis

Time per task

seconds per task · #25 of 26
  1. 723723Gemini 4 Argon (High)
  2. 736736Claude Fable 5.1 (Max)
  3. 791791Claude Opus 5.5 (Max)
  4. 824824GLM-5.3-Flash
  5. 846846Kimi K3.1 Pro
  6. 901901Grok 4.7 (Xhigh)
  7. 936936Claude Sonnet 5.5 (Max)
  8. 1,0131,013MiMo-V2.6-Pro
  9. 1,2001,200Qwen3.8 27B (Xhigh)
  10. 2,0572,057Qwen3.8 Max (0902)
Data from Artificial Analysis

Tokens per task

output tokens per task · #15 of 27
  1. 54.5K55KMuse Wonder (Max)
  2. 55.2K55KKimi K3.1 Pro
  3. 61.6K62KGemini 4 Argon (High)
  4. 64.1K64KStep 5 Preview
  5. 64.3K64KMiMo-V2.6-Pro
  6. 66.8K67KQwen3.8 27B (Xhigh)
  7. 68.7K69KGLM-5.3-Flash
  8. 69.5K70KMistral Large 4 Preview
  9. 71K71KGemini 3.8 Flash (High)
  10. 71.1K71KGLM-5.3 (Max)
Data from Artificial Analysis

Speed and price

Speed

tokens per second · #151 of 162
  1. 4949Qwen3.8 27B (Low)
  2. 4848Qwen3.8 27B (Medium)
  3. 4848Granite 4.2 30B
  4. 4646Phi-4 Mini Instruct
  5. 4545Gemma 4 E4B (Reasoning)
  6. 4545Qwen3.8 27B (Xhigh)
  7. 4545Gemma 4 E4B (Non-reasoning)
  8. 4343Hermes 4 - Llama-3.1 405B (Non-reasoning)
  9. 4343Hermes 4 - Llama-3.1 405B (Reasoning)
  10. 4141Qwen3.8 2.4T A95B
Data from Artificial Analysis

Latency

seconds to first token · #114 of 162
  1. 3.13.1MiMo-V2.6-Pro
  2. 3.53.5Apertus 8B Instruct
  3. 3.63.6Mercury 2.5
  4. 3.73.7Ring-2.6-1T
  5. 3.83.8MiMo-V2.6-Flash
  6. 3.93.9Qwen3.8 27B (Xhigh)
  7. 3.93.9Qwen3.8 27B (Low)
  8. 44Qwen3.8 27B (Non-reasoning)
  9. 44Qwen3.8 27B (Medium)
  10. 44GPT-6.1 Sol (Medium)
Data from Artificial Analysis

Price

$ per 1M tokens (3:1) · #99 of 169
  1. $0.850.85Nova 2.0 Lite (Low)
  2. $0.850.85Nova 2.0 Lite (Non-reasoning)
  3. $0.980.98Apodex 1.1
  4. $1.11.1Qwen3.5 122B A10B (Non-reasoning)
  5. $1.11.1Qwen3.5 122B A10B (Reasoning)
  6. $1.131.13Qwen3.8 27B (Xhigh)
  7. $1.131.13Qwen3.8 27B (Medium)
  8. $1.131.13Qwen3.8 27B (Low)
  9. $1.131.13Nemotron 3 Ultra 550B A55B (Reasoning)
  10. $1.131.13Qwen3.8 27B (Non-reasoning)
Data from Artificial Analysis

Nearby models

  • GPT-6 Luna (Xhigh)OpenAI34.6
  • Claude Haiku 5.5 (Medium)Anthropic34.5
  • GLM-5.3 (Low)Z AI34.3
  • GPT-5.6 Terra (High)OpenAI34.2
  • JT-4.1 Flash 236B A21B (Reasoning)China Mobile33.9
  • Motif 3Motif Technologies33.6
  • Gemini 3.8 Flash (Low)Google33.5
  • GPT-6 Luna (High)OpenAI32.9
Data from Artificial Analysis

Same lab

  • Qwen3.8 Max (0902)Alibaba45.4
  • Qwen3.8 2.4T A95BAlibaba39.9
  • Qwen3.8-Flash-NextAlibaba39.8
  • Qwen3.8 27B (Medium)Alibaba27.6
  • Qwen3.8 27B (Low)Alibaba26.2
  • Qwen3.7 PlusAlibaba25.2
  • Qwen3.5 397B A17B (Non-reasoning)Alibaba21.4
  • Qwen3.5 Omni PlusAlibaba20.4
Data from Artificial Analysis
HorizonVascend IDEOlympusTalosCoursesCommunityBenchmark
DiscordGitHubXLet’s talkLegalPrivacyCookiesData useSecurity
© 2026 Vascend Technologies · VAT IT04911550616
ascend