Compare Models
Gemini 3.8 Flash

Release Date: Sep 2, 2026

Developer Google 🇺🇸
Context Window 1M
Max Output Tokens 66k
Token Costs (in/out) $1.50/7.50
Weights Private
Input Modalities

Accuracy

62.25 % ± 1.01

Cost / Test (Vals Index)

$ 5.392

Latency

44 min 53 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

0.0%

±1.01
7/57
RSI Index

0.0%

7/10
Code Migration

0.0%

±4.18
21/59
CyberBench

0.0%

±2.21
22/25
EMB

0.0%

±2.42
5/57
Finance Agent (v2)

0.0%

±0.13
1/60
Time Horizon Index: KSP

0.0%

±0.00
4/11
Legal Research Bench

0.0%

±3.39
20/60
MedCode

0.0%

±2.18
24/92
MedScribe

0.0%

±1.94
26/94
MortgageTax

0.0%

±0.94
44/98
ProofBench v1.1

0.0%

±5.02
17/31
SAGE

0.0%

±3.36
59/81
Public Benefits Bench v1.1

0.0%

±1.24
12/35
Tax Agent Bench

0.0%

±3.15
8/20
TaxEval v2

0.0%

±0.85
36/145
Vibe Code Bench v1.1

0.0%

±3.88
14/95
BioMysteryBench

0.0%

±1.70
9/11
Harvey's Legal Agent Benchmark

0.0%

±2.42
11/61
GPQA Diamond

0.0%

±1.48
4/138
IOI

0.0%

±2.26
10/27
LiveCodeBench

0.0%

±0.90
3/143
LegalBench

0.0%

±0.43
6/144
MMLU Pro

0.0%

±0.29
5/138
MMMU Pro

0.0%

±0.75
4/93
ProgramBench

0.0%

±0.70
9/45
SkillsBench

0.0%

±4.42
12/35
SWE-bench

0.0%

±1.79
24/88
Terminal-Bench 2.1

0.0%

±0.38
5/65
Terminal-Bench Science

0.0%

±3.37
9/25
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Google
Temperature: 1
Top P: Default
Top K: Default
Max Output Tokens: 65,536
Reasoning Effort: high