Nemotron 3.5 Bolt (74B A9B 24GB)

Multimodal

NVIDIA

Nemotron 3.5 Bolt is NVIDIA's open multimodal 74B total / 9B active mixture-of-experts vision-language model, compressed from Nemotron 3.5 Super VL into a native 2-bit 24GB footprint for RTX and DGX Spark class machines. It takes text, image and video input with up to 256K context and configurable reasoning. Released 2026-10-15 under OpenMDW-1.1, with a full-precision BF16 sibling published alongside.

Key Specifications

Parameters

74.0B

Context

-

Release Date

October 15, 2026

Average Score

50.8%

API Documentation Research Paper Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

October 15, 2026

Last Update

October 20, 2026

Today

October 27, 2026

Technical Specifications

Parameters

74.0B

Training Tokens

-

Knowledge Cutoff

May 1, 2026

Family

-

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Reasoning

Logical reasoning and analysis

GPQA

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card; GPQA Diamond • Self-reported

86.4%

Other Tests

Specialized benchmarks

AA-LCR

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

64.1%

BrowseComp

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

66.8%

Humanity's Last Exam

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card; HLE (no tools) • Self-reported

25.3%

MCP Atlas

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

52.2%

OmniScience

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card; AA-Omniscience accuracy • Self-reported

16.9%

PinchBench

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

87.8%

SciCode

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

41.0%

Tau3 Banking

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card; τ³-bench (Banking) • Self-reported

17.0%

Terminal-Bench 2.1

NeMo Gym / NeMo Evaluator; IQ2 24GB column as on HF card • Self-reported

51.0%

License & Metadata

License

openmdw

Announcement Date

October 15, 2026

Last Updated

October 20, 2026

Similar Models

Nemotron 3.5 Super VL (120B A12B)

NVIDIA

Best score:0.9 (GPQA)

Released:Oct 2026

Price:$0.20/1M tokens

Gemma 4 12B

Google

Best score:0.8 (GPQA)

Released:May 2026

Gemma 4 31B

Google

Best score:0.8 (GPQA)

Released:Apr 2026

Price:$0.39/1M tokens

Gemma 4 26B-A4B

Google

Best score:0.8 (GPQA)

Released:Apr 2026

Price:$0.13/1M tokens

Muse Glimmer-30B

Meta

Best score:0.8 (GPQA)

Released:Aug 2026

Qwen3 VL 32B Thinking

Alibaba

Best score:0.9 (MMLU)

Released:Sep 2025

Qwen3.6-35B-A3B

Alibaba

Best score:0.9 (GPQA)

Released:Apr 2026

Qwen3.8-27B

Alibaba

Best score:0.9 (GPQA)

Released:Aug 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.