Declare Lab (SUTD)
NORA
3.76B Commercial terms unverified 7/8 verified
1 of 8 required fields are unverified. They are shown as badges, not filled with defaults. Sources for verified fields are listed below.
Facts
- Released
- 2025-04 1 Source
- Parameters
- 3.76 B 1 Source
- Architecture
- Monolithic autoregressive 0 Source
- License
- Neither repository nor model card has a licence tag 0 Source
- Commercial
- Commercial terms unverified 0 Source
- Weights
- huggingface.co ↗ 1 Source
- Code
- github.com ↗ 1 Source
- Paper
- arxiv.org ↗ 0 Source
- Input modalities
- rgb · language 0 Source
- Action space
- 7-DoF (6 + gripper) 0 Source
- Cross-embodiment
- Open X-Embodiment mixture pretraining; WidowX / BridgeData V2 evaluation 1 Source
- Pretraining data (hours)
- Unverified 0 Source
- Target forms
- Single arm 0 Source
Architecture · Qwen2.5-VL-3B backbone, autoregressive action tokens; NORA-1.5 and NORA-LONG variants
Sources · 2 sources
arch_notes,cross_embodiment,code_url https://github.com/declare-lab/nora Official page · Fetched 2026-09-04 · No snapshot stored yet README：built on Qwen2.5 VL；OXE mixture；WidowX / BridgeData V2；7-DoF
params_b,weights_url,release_date https://huggingface.co/api/models/declare-lab/nora Model hub page · Fetched 2026-09-04 · No snapshot stored yet safetensors 3,758,262,272; model card has no license field; createdAt 2025-04-28
Activity signals
Public counters fetched daily from GitHub, Hugging Face and ModelScope. Numbers only; no judgement. · fetched 2026-10-31
GitHub stars
225
+1 / 30 d
Commits · 90 d
0
last push 296 d ago
HF downloads · 30 d
197
all-time 39,981 · 13 likes
Weights last modified
2025-08-27
431 d ago
Hugging Facedeclare-lab/nora
stars · last 30 d
Public benchmark scores
As printed in the source; protocols differ between papers. Full table with notes on the benchmarks page.
| Benchmark | Subset | Score | Source |
|---|---|---|---|
| LIBERONORA-fine-tuned | avg | 73.9 | paper · Table 2, row 'NORA-fine-tuned (Ours)', Average |
| LIBERONORA-fine-tuned-AC | avg | 79.5 | paper · Table 2, row 'NORA-fine-tuned-AC (Ours)', Average |
| LIBERONORA-Long-fine-tuned | avg | 87.9 | paper · Table 2, row 'NORA-Long-fine-tuned (Ours)', LIBERO-Average |
| LIBERONORA-Long-fine-tuned | goal | 89.4 | paper · Table 2, row 'NORA-Long-fine-tuned (Ours)', LIBERO-Goal |
| LIBERONORA-Long-fine-tuned | long | 74.6 | paper · Table 2, row 'NORA-Long-fine-tuned (Ours)', LIBERO-Long |
| LIBERONORA-Long-fine-tuned | object | 95.4 | paper · Table 2, row 'NORA-Long-fine-tuned (Ours)', LIBERO-Object |
| LIBERONORA-Long-fine-tuned | spatial | 92.2 | paper · Table 2, row 'NORA-Long-fine-tuned (Ours)', LIBERO-Spatial |
| Real-robot evaluation (self-reported) | avg | 56.7 | paper · Table 1, row 'Average', column 'NORA (Ours)' |
Deployment recipe
Environment, weights, run command and the pitfalls we hit, exactly as executed on rented GPUs (venv qwen, 7.5 GB of weights, 2 pitfalls).
Open recipe →Record history
2026-09-04
Sources fetched
arch_notes,cross_embodiment,code_url · params_b,weights_url,release_date
Latency & cost
cost = p50 ms ÷ 1000 × rental ÷ 3600 × 1000
| Hardware | p50 ms | p95 ms | Peak VRAM | Cost per 1k inferences ¥ | samples |
|---|---|---|---|---|---|
NVIDIA RTX 4090 ¥1.68/h · GongJi Compute (2026) | 112 | 127 | 7.13 GB | 0.052 | 1 Preliminary (n<3) Maintainer measured |
NVIDIA RTX 5090 ¥2.4/h · Runyun 2.4 / GongJi 2.5 / Paratera Cloud 2.98 (2026) | 129 | 135 | 7.13 GB | 0.086 | 1 Preliminary (n<3) Maintainer measured |
NVIDIA A100 80GB no rental price | — | — | — | — no latency data | No data |
NVIDIA Jetson AGX Thor no rental price | — | — | — | — no latency data | No data |
NVIDIA H100 80GB no rental price | — | — | — | — no latency data | No data |
NVIDIA A800 80GB no rental price | 146 | 151 | 7.13 GB | — no rental price for this hardware | 1 Preliminary (n<3) Maintainer measured |
Compatibility
Colored only where we hold evidence. An amber dot means the vendor declares this form as a target; that is a claim, not a verified compatibility.
| Unitree G1 | Unitree H1 | AgiBot Genie G1 | AgileX Piper arm | SO-101 | Franka FR3 / Panda | Fourier GR-2 | WidowX 250 S | Google Robot（Everyday Robots） | ALOHA 2 / Cobot Magic (dual-arm) | Galaxea R1 Pro | Franka Emika Panda | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NORA |
Legend Officially supportedCommunity reproducedFine-tunable in theoryUnsupportedUnknown Vendor-declared target form 1 / 12 cells with a definite status
Reproduction records
Did this model run on your robot? Report it with a link (log, video or repo). Records are reviewed by the maintainers; a verified “works” or “works after fine-tuning” upgrades the matrix cell to “community verified” with your link as evidence. Anonymous is fine; nothing but the fields below is stored.
No verified record yet for this model.
Clear direction, grounded judgement.