Inference
EN

Ling 2.6 Flash vs Nemotron 3.5 Lightning 30B A3B NVFP4

Change models
Ling 2.6 FlashInclusionAI
Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA
IndexAll categories, out of 100.
37.0 ±7.9
37.0 ±6.8
GapIndex points below the top.
Top
Top
CoverageIndex weight that has results.
75%
75%
SpeedOutput tokens per second.
104/s
—
EloLMArena rating and rank.
—
—
ReleasedMonth of release.
2026-04
2026-08
Input / 1MUS dollars per 1M input tokens.
—
—
Output / 1MUS dollars per 1M output tokens.
—
—
1K chats2K tokens in, 500 out, each.
—
—
1K documents20K tokens in, 1K out, each.
—
—
1K agent tasks200K tokens in, 20K out, each.
—
—
ContextMaximum tokens in one request.
262K
1M best
AgenticMulti-step tasks with tools.
46.9 best
30.2
CodingCode writing and repair.
31.0
38.7 best
ReasoningLogic problems and puzzles.
32.8
38.9 best
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
30.8
39.5 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
42.5
57.1 best
MathMath problems.
—
—

Benchmarks

9 shared
Only Ling 2.6 Flash: 2Only Nemotron 3.5 Lightning 30B A3B NVFP4: 10
Ling 2.6 Flash
Nemotron 3.5 Lightning 30B A3B NVFP4
Artificial Analysis Humanity's Last ExamKnowledge
6.3%168 / 222
10.6% best143 / 222
Artificial Analysis Long Context ReasoningReasoning
31.3%179 / 217
49.2% best159 / 217
Critical Physics TasksReasoning
0.0%150 / 214
0.0%150 / 214
Artificial Analysis GPQA DiamondKnowledge
59.3%157 / 199
74.3% best122 / 199
Artificial Analysis Omniscience AccuracyKnowledge
15.6% best149 / 164
14.7%153 / 164
Graduate-Level Google-Proof Q&AKnowledge
59.0%113 / 141
75.6% best78 / 141
GDPval-AA normalizedAgentic
0.0%101 / 133
6.9% best93 / 133
Instruction Following BenchmarkInstruction
57.0%46 / 57
72.9% best37 / 57
Scientific Code BenchmarkCoding
27.0%33 / 35
31.4% best30 / 35