AgentTested hands-on review
tool: NVIDIA Nemotron 3.5 Super
slug: nemotron-3-5-super
date: 2026-10-21
score: 8/10

Coverage: Same 4-task battery as MiniMax M3.1 Flash, 2 runs, through OpenRouter's free route
What we tested: Same 4-task battery as MiniMax M3.1 Flash (code with 6 unit checks, 2 trick questions, 60-word summary of a 700-word text), 2 runs.
Method: OpenRouter nvidia/nemotron-3.5-super-vl:free, temperature 0, script _battery/battery.py. Cost reported 0.

Notes:
- CPU-only box: 8 cores, 15 GB RAM, no GPU.
- Hosted; free daily limit

Results:
pass_rate: 8/8
latency: 0.9-5.5 s
run1.code: passed, 6/6 unit cases, wall 1.36 s, 173 out tokens
run1.reason1: passed, expect 5, wall 1.55 s, 117 out tokens
run1.reason2: passed, expect 5, wall 0.88 s, 100 out tokens
run1.summary: passed, 52 words, 3/3 key terms, wall 5.3 s, 684 out tokens
run2.code: passed, 6/6 unit cases, wall 1.9 s, 274 out tokens
run2.reason1: passed, expect 5, wall 1.13 s, 117 out tokens
run2.reason2: passed, expect 5, wall 1.19 s, 105 out tokens
run2.summary: passed, 59 words, 3/3 key terms, wall 5.54 s, 993 out tokens
cost_usd: 0

Pros:
- 8/8 task runs passed, the only perfect score in our chat battery so far
- Fast: short answers in 1-2 s, summaries in about 5 s

Cons:
- Spends many hidden reasoning tokens (up to 896 on a summary)
- Free tier has a daily request cap across all free models

Verdict: Nemotron 3.5 Super was the best free chat model we've run. It passed every task in both runs, including keeping the summary under 60 words, which MiniMax missed once, and it was quick. This is the same small battery as before, so it is a strong but narrow result.

Score line: 8/8 on a small battery, fast, $0.