Model
Model explorer

Nemotron 3.5 Bolt E4B

OPEN
NVIDIA · Nemotron 3.5 family · released Oct 23, 2026

Second model in NVIDIA's edge-focused Bolt line, released Oct 23, 2026 on Hugging Face and build.nvidia.com as announced with Bolt 74B-A9B on Oct 15: a Gemma-4-E4B post-trained by NVIDIA for fast local voice and device assistants, personas and embedded interfaces, with stronger multi-turn tool use, instruction following and input grounding. It ships as a 4GB Q4_K_M build plus BF16 weights under OpenMDW-1.1. NOT FILED: Tau2-Bench (55.10 BF16 / 52.32 Q4; the card links the aggregate 'core' board, while the corpus has per-domain tau2 columns), and MultiChallenge, RoleBench, RoleMRC, EQ-Bench v3, Brevity, When2Call, ToolSandbox, IHEval, AgentIF, StructEval-T, InverseIFEval, RULER (64K), LongMemEval, LongBench v2, SysBench, GaRAGe, RAGTruth and RelyToolBench, which have no corpus column.

ReasoningCodingVisionFunction callingTool useAgentic
1569.2
Elo · unrated
Parameters
8B
Active params
8B (dense)
Context
64K tokens
Architecture
Gemma-4-E4B architecture (8B total, 4B effective parameters), post-trained by NVIDIA from Gemma-4-E4B-it with multi-environment GRPO RL; 64K-token context; text in, text out; Q4_K_M 4GB release with BF16 reference weights and a DFlash speculative-decoding drafter
License
OpenMDW License 1.1 (base Gemma 4 E4B under Apache-2.0)
Languages
—
API price (in/out)
No hosted API
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
BFCL v3Agents67.3%#17
best: Hunyuan-A13B · 78.3%
IFBenchReasoning64.0%#37
best: MiniMax M3 · 83.0%
Run it locally
VRAM @ Q4
6.18 GB
VRAM @ FP16
—
Fits on (Q4)
RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB
Throughput data unavailable.
Quantizations
4-bit · GGUF
Fine-tune it
Permissive
QLoRA7.0 GB1× RTX 3060 12GB
LoRA18.6 GB1× RTX 3090 24GB
Full fine-tune130.0 GB1× H200 141GB
QLoRA SFT on ~10k samples ≈ $4.68 (1× RTX 3060 12GB)
Plan a fine-tune →
Nemotron 3.5 family
Elo progression across releases
Nemotron 3.5 Bolt 74B-A9BOct 20261952.2Nemotron 3.5 Super VLOct 20262076.9Nemotron 3.5 Bolt E4BOct 20261569.2
Resources
HFnvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GBDOCShttps://research.nvidia.com/labs/nemotron/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB/APIAvailable via NVIDIA API
API price weights · each benchmark row carries its own source badge (see methodology)