Compare Models

Qwen 3.8 27B

Qwen 3.8 27B is an open-weight model from Alibaba, released August 14, 2026. Its best result is #7 of 90 on SAGE.

Release Date: Aug 14, 2026

Developer Alibaba  🇨🇳
Context Window 262k
Max Output Tokens 66k
Token Costs (in/out) $0.50/3.00
Weights Open
Input Modalities

Accuracy

52.97 %

Avg. Cost (In/Out)

$ 0.50 / $ 3.00

Latency

49 min 36 s

Vals Index
BenchmarksAccuracyRankings
Code Migration

14.16%

±4.19
54/74
CyberBench v1.1

67.80%

±5.68
25/44
EMB

59.66%

±3.03
33/71
Finance Agent (v2)

48.55%

±2.10
44/75
Legal Research Bench

36.06%

±3.34
36/74
MedCode

28.70%

±1.97
97/105
MedScribe

83.85%

±1.98
39/107
MortgageTax

64.94%

±0.95
45/98
ProofBench v1.1

16.00%

±3.69
41/47
SAGE

52.40%

±3.47
7/90
Tax Agent Bench

60.04%

±3.28
33/66
TaxEval v2

70.85%

±0.89
88/145
Vibe Code Bench v1.1

64.85%

±4.83
41/109
Harvey's Legal Agent Benchmark

11.25%

±2.42
12/75
GPQA Diamond

88.89%

±1.58
33/138
IOI

39.06%

±4.94
34/41
LiveCodeBench

84.00%

±1.03
43/143
LegalBench

82.43%

±0.47
70/149
MMLU Pro

84.34%

±0.36
65/138
MMMU Pro

83.93%

±0.88
28/93
ProgramBench

0.00%

±0.00
48/62
SkillsBench

38.11%

±4.38
31/35
SWE-bench

86.00%

±1.55
15/88
Terminal-Bench 2.1

58.43%

±5.25
42/75
Terminal-Bench Science

0.00%

±0.00
38/38
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Alibaba
Temperature: 1
Top P: 0.95
Top K: 20
Max Output Tokens: 65,536
Reasoning Effort: xhigh