#295

Nemotron 3.5 Lightning

NVIDIA Release: 2026-08-11 Tested on: 2026-08-14 03:49 nvidia/nemotron-3.5-lightning::medium

30B total (3B active)MoEWeights available

(high) (medium) (low) (none)

Summary

Nemotron 3.5 Lightning scores 5.1 on AI BENCHY and ranks #295. It has 9.9 reliability, a 48.5% pass rate, $0.156 total cost, and 70.11s average response time.

What makes Nemotron 3.5 Lightning unique: It stands out most in Instructions following, where it ranks #1, while Data parsing and extraction is its weakest area at #15. It uses unusually many reasoning tokens, which can help explain its slower or more expensive runs.

Model facts

Researched on 2026-08-12

Reported

Parameters: 30B total (3B active)
Architecture: MoE
Availability: Weights available
License: NVIDIA Open Model

Parameter source Availability source

Score

5.1

Consistency

6.0

Reliability

9.9

Total Cost (Current Price)

$0.156 ↕ New

Tested at: $0.000

Total Output Tokens

760,441

Total Input Tokens

126,330

Input Price

$0.080 / 1M

Output Price

$0.200 / 1M

Tests Correct

Wrong Tests: 17

Attempt pass rate: 48.5%

Flaky tests

Flaky tests had mixed outcomes across runs (at least one pass and one fail).

Response Time (avg)

70.11s

Response Time (max): 437.11s

Response Time (total): 1542.52s

Wrong answer: 13 Did not follow instructions: 3 Invalid tool call: 1

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#295 Nemotron 3.5 Lightning

medium

Cost: $0.000
Time: 130.5s
Tokens: 6,750 tok

Run history

Tested on	Score	Reliability	Tests Correct	Total Cost	Compare
2026-08-14 03:49 Re-test	5.1	9.9		$0.156