Inference
EN

GPT-5.4 nano vs Nemotron 3.5 Super VL 120B A12B

Change models
GPT-5.4 nanoOpenAI
Nemotron 3.5 Super VL 120B A12BNVIDIA
IndexAll categories, out of 100.
50.9 ±2.5
51.0 ±5.1 best
GapIndex points below the top.
−0.1 · no clear gap
Top
CoverageIndex weight that has results.
100% best
85%
SpeedOutput tokens per second.
91/s
—
EloLMArena rating and rank.
1372
—
ReleasedMonth of release.
2026-03
2026-10
Input / 1MUS dollars per 1M input tokens.
$0.20
—
Output / 1MUS dollars per 1M output tokens.
$1.25
—
Batch output / 1MBatch price per 1M output tokens.
$0.625
—
1K chats2K tokens in, 500 out, each.
$1.03
—
1K documents20K tokens in, 1K out, each.
$5.25
—
1K agent tasks200K tokens in, 20K out, each.
$65.00
—
ContextMaximum tokens in one request.
400K
1M best
AgenticMulti-step tasks with tools.
50.5 best
47.8
CodingCode writing and repair.
53.7 best
53.0
ReasoningLogic problems and puzzles.
53.7
56.1 best
MultimodalTasks with images and text.
41.7
47.4 best
KnowledgeFacts and expert knowledge.
44.5
52.7 best
MultilingualTasks in many languages.
44.4
—
InstructionTasks with strict prompt rules.
61.3 best
55.2
MathMath problems.
55.9
—

Benchmarks

11 shared
Only GPT-5.4 nano: 63Only Nemotron 3.5 Super VL 120B A12B: 17
GPT-5.4 nano
Nemotron 3.5 Super VL 120B A12B
Artificial Analysis Humanity's Last ExamKnowledge
28.3%90 / 222
30.7% best74 / 222
Artificial Analysis Long Context ReasoningReasoning
76.7% best84 / 217
68.6%121 / 217
Critical Physics TasksReasoning
9.3%69 / 214
10.9% best62 / 214
Artificial Analysis Omniscience AccuracyKnowledge
25.7%102 / 164
26.2% best97 / 164
GDPval-AA normalizedAgentic
22.4%79 / 133
33.1% best55 / 133
Vibe Code Bench v1.1Coding
26.1% best83 / 123
3.7%111 / 123
Artificial Analysis SciCodeCoding
47.2%64 / 114
47.8% best61 / 114
Legal Research BenchLegal
6.3%84 / 90
10.1% best81 / 90
Tax Agent BenchFinance
5.0%78 / 82
11.6% best65 / 82
Humanity's Last ExamKnowledge
37.7% best36 / 77
29.2%49 / 77
Humanity's Last Exam without toolsKnowledge
24.3%47 / 53
31.8% best35 / 53