Holotron-12B
Holotron-12B is available now for agents, long context, and vision with open-weight and 128K context; evaluate it while provider pricing coverage matures.
- Family
- Holotron
- Released
- 2026-03-16
- Context
- 128K
- Parameters
- 12
- Architecture
- SSM Hybrid
- Specialization
- computer-use
- Openness
- Open weights
- License
- NVIDIA Open ModelPermitted
- Weights
- Available
- Code
- Unknown
- Training
- Supervised Fine-tuning
No tracked provider pricing is available yet.
About
Holotron-12B is H Company's 12B-parameter computer-use VLM released March 16, 2026, built on NVIDIA's Nemotron-Nano-12B-v2-VL with a Hybrid SSM+Attention architecture (Mamba-2 layers). Achieves 80.5% on WebVoyager and ~8.9k tokens/s throughput at concurrency 100 - over 2x faster than Holo2-8B. Licensed under NVIDIA Open Model License.
Overview
Holotron-12B is H Company's 12B-parameter computer-use VLM released March 16, 2026, built on NVIDIA's Nemotron-Nano-12B-v2-VL with a Hybrid SSM+Attention architecture (Mamba-2 layers). Achieves 80.5% on WebVoyager and ~8.9k tokens/s throughput at concurrency 100 - over 2x faster than Holo2-8B. Licensed under NVIDIA Open Model License.
Providers
No active provider routes are tracked for this model yet.
Capabilities
Benchmark peer barsfor Agents
No task-mapped benchmark peers are available for this model yet.
Benchmark scores(1)
| Benchmark | Score | Version | Evaluation | Source |
|---|---|---|---|---|
| 80.5 | WebVoyager - navigation agentObserved 2026-03-16 | — |
Migration checks
No linked migration route is available for this model yet.
No tracked provider pricing is available yet.