Nemotron 3.5 Bolt E4B

Nemotron 3.5 Bolt E4B.

NVIDIA Released October 23rd, 2026 Family Nemotron 3.5 The lab's model card Compare with others

Benchmarks with a score

Ranked first, out of five or more models

In the top three

NVIDIA Oct 2026

Nemotron 3.5 Bolt E4B

Open weights Small

128.3 #286 of 404

Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.

MMMU-Pro 42% #83 /90
IFBench 64% #54 /80
MultiChallenge 56% #23 /42

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
AgentIF (2026-10 setting)	64.7%	#7 of 24	Official leaderboard
MultiChallenge	56%	#23 of 42	Official leaderboard
EQ-Bench 3	1,199	#43 of 78	Official leaderboard
RULER	85.8%	#28 of 42	Lab model card
IFBench	64%	#54 of 80	Lab model card
Creative Writing v3	57.0	#92 of 120	Official leaderboard
LongBench v2	48.7%	#27 of 35	Lab model card
MMLU-ProX	66.2%	#19 of 23	Lab model card
SysBench	67.2%	#6 of 6	Lab model card
Inverse IFEval	45.3%	#3 of 3	Lab model card
AgentIF	22.9%	#1 of 1	Lab model card
Brevity Benchmark	82.8%	#1 of 1	Lab model card
GaRAGe	40.7%	#1 of 1	Lab model card
LongMemEval	69.9%	#1 of 1	Lab model card
RAGTruth	77.9%	#1 of 1	Lab model card
RoleBench	71.2	#1 of 1	Lab model card
RoleMRC	79.1%	#1 of 1	Lab model card

Coding

Benchmark	Score		Rank	Source
StructEval	80%		#1 of 1	Lab model card

Agents and tools

Benchmark	Score	Rank	Source
Berkeley Function Calling Leaderboard (BFCL)	67.3%	#24 of 41	Lab model card
ToolSandbox	74.1%	#4 of 5	Lab model card
tau2-bench	52.3%	#71 of 79	Lab model card
RelyToolBench	69.3%	#1 of 1	Lab model card
When2Call	60.5%	#1 of 1	Lab model card

Professional work

Benchmark	Score		Rank	Source
MedXpertQA-MM	26.9%		#19 of 21	Lab model card

Vision, audio and robotics

Benchmark	Score	Rank	Source
VoiceBench	80.3%	#10 of 12	Lab model card
CoVoST2	36.1	#12 of 14	Lab model card
MATH-Vision (MathVision)	46.3%	#46 of 51	Lab model card
MMMU-Pro	42%	#83 of 90	Lab model card
Audio MC	25.2%	#6 of 6	Lab model card

Safety and security

Benchmark	Score		Rank	Source
IHEval	82.4%		#1 of 2	Lab model card

Its lab's numbers put it at 128.3, number 286 of 404

The Self-Reported Index fits one score to the 11 capability results NVIDIA reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 125.2 to 131.5. How we compute it.

RULER 85.8
VoiceBench 80.3
Berkeley Function Calling Leaderboard (BFCL) 67.3
MMLU-ProX 66.2
IFBench 64
MultiChallenge 55.6
tau2-bench 52.3
LongBench v2 48.7
MATH-Vision (MathVision) 46.3
MMMU-Pro 42
MedXpertQA-MM 26.9

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap
MultiChallenge	55.6% MultiChallenge	56% Leaderboard, thinking	−0.4 pts

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

NVIDIA-Nemotron-3.5-Bolt-E4B-4GB model card (Hugging Face) Oct 23, 2026

Its system card

Every result in the longer documents the lab published with it. 11 of them aren't in the launch post.

Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents Oct 23, 2026