Gemini 4 Neon

Gemini 4 Neon is a model from Google, released November 18, 2026. It ranks #2 of 42 models on the Vals Index with 70.87%. Its best result is #1 of 99 on Legal Research Bench.

Release Date: Nov 18, 2026

Developer Google ย ๐Ÿ‡บ๐Ÿ‡ธ
Context Window 1M
Max Output Tokens 1M
Token Costs (in/out) $3.00/15.00
Weights Private
Input Modalities

Accuracy

70.87 % ยฑ 0.94

Cost / Test (Vals Index)

$ 7.405

Latency

31 min 25 s

Vals Index
BenchmarksAccuracyRankings

70.87%

ยฑ0.94
2/42

68.31%

ยฑ1.38
1/45

78.36%

ยฑ1.83
1/43

62.63%

ยฑ2.33
4/28

76.46%

ยฑ2.06
4/47

68.56%

ยฑ1.94
2/45

56.25%

ยฑ3.45
1/99

45.05%

ยฑ3.35
9/34

54.89%

ยฑ3.59
1/90

93.70%

ยฑ1.17
2/133

80.00%

ยฑ1.93
4/34

19.58%

ยฑ3.31
8/43

61.62%

ยฑ2.02
7/69

62.86%

ยฑ5.82
2/48
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Google
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 1,048,576
Reasoning Effort: xhigh

Updates

Nov 18, 2026

We evaluated Googleโ€™s new Gemini 4 Neon across our benchmark suite.

Neon is on the scoreโ€“cost Pareto frontier on 7 of its 14 leaderboards: the Vals, Finance, and Legal indexes, Earnings Modeling Bench v1.1 ($3.82/test), Legal Research Bench ($7.37/test), Tax Agent Bench ($4.95/test), and Vibe Code Bench ($6.76/test). On those boards, no model with a reported cost scores at least as well for less or scores better for the same cost.

Pricing is $3 per million input tokens, $15 per million output tokens, and $0.15 per million cache-read tokens.