System Card← leaderboard
Alibaba

Qwen Lite

#25of 28 · general

Qwen3.8 27B, open weights

28Bparams
262kcontext
$0.50 / $3per 1M in / out
2026-08-14released
weights ↗repo ↗
50100Coding 20 · #25 of 28 · 44% of the recipeAgents 24 · #24 of 29Cyber 3 · #20 of 24Engineering 2 · #26 of 26Math 12 · #24 of 26Health 13 · #22 of 24Legal 17 · #25 of 27Professional 11 · #26 of 28Writing 4 · #19 of 21Instruct 19 · #20 of 25Hallucination 19 · #21 of 26 · 47% of the recipeMultimodal 9 · #23 of 24
Coding20#25
Agents24#24
Cyber3#20
Engineering2#26
Math12#24
Health13#22
Legal17#25
Professional11#26
Writing4#19
Instruct19#20
Hallucination19#21
Multimodal9#23

History

ECI144146148150152154Jul 2026Qwen3.6 27B · $0.25/M in · 2026-04-22 · still served · Epoch Capabilities Index 146.2 ECI pointsQwen3.8 27B · $0.02/M in · 2026-08-14 · still served · Epoch Capabilities Index 151.7 ECI pointsn3.6 27B
  • Alibaba
  • older → newer
$ / M input tokens2 of 176 priced models · price is a third-party claim, not a vendor quote

Effort

Qwen3.8 27B
score1520253035$1$1.5$2$3$ / task · log scaleNoneLowMedXhigh

Measured

≈ imputed · amber dot = lab-self-reported among the sources
GPQA Diamond89.0%#19LLM StatsVals.ai
HLE32.4%#27Artificial AnalysisLLM Stats
MMLU-Pro84.3%#8Vals.ai
LiveBench75.3%#15LiveBench
SWE-bench V86.0%#4Vals.ai
LiveCodeBench84.0%#6Vals.ai
SciCode46.6%#26Artificial AnalysisEpoch AI
CritPt5.4%#26Artificial AnalysisEpoch AI
Terminal-Bench58.4%#16Vals.ai
APEX Agents47.5%#15Epoch AI
GDPval46.0%#22Artificial Analysis
Arena1438#20Arena
AA Index33.7#23Artificial Analysis
Epoch Capabilities Index151.7#18Epoch AI

Neighbors

Mistral→ stronger math · pricierMiniMax→ weaker agent · cheaperKimi Code→ narrower · pricierStep→ more factualMiMo Flash→ cheaper