← All models

Nemotron 3.5 Bolt E4B BF16

NVIDIA · 2026-10-23 · 8B parameters

Nemotron 3.5 Bolt E4B BF16 is the full-precision release of NVIDIA's Nemotron 3.5 Bolt E4B, a small text model post-trained by NVIDIA on Google's Gemma 4 E4B (about 8B total and 4B effective parameters; the checkpoint holds 7,996,156,490) for local voice and device assistants, personas and embedded interfaces. It supports up to 64K input and 64K output tokens, switchable thinking with a runtime thinking budget, and a conversational mode. NVIDIA also releases a 4GB Q4_K_M build with its own scores. The weights are under OpenMDW-1.1, with the Gemma 4 base under Apache 2.0.

Benchmark scores