Stanford ILIAD
MiniVLA
500M 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
- 2024-12 1 Source
- Parameters
- 0.5 B 1 Source
- Architecture
- Monolithic autoregressive 0 Source
- License
- Code MIT; README notes pretrained weights may inherit base-model (Llama Community License) restrictions 1 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 end-effector pose delta + gripper (discrete tokens) 0 Source
- Cross-embodiment
- Open X-Embodiment mixture; four LIBERO task suites; WidowX / BridgeData V2 1 Source
- Pretraining data (hours)
- Unverified 0 Source
- Target forms
- Single arm 0 Source
Architecture · Scaled-down OpenVLA: Prismatic architecture + Qwen2.5-0.5B language backbone + DINOSigLIP vision, 256 extra action tokens
Sources · 1 sources
release_date,params_b,license,arch_notes,cross_embodiment,code_url,weights_url https://github.com/Stanford-ILIAD/openvla-mini Official page · Fetched 2026-09-05 · No snapshot stored yet README：2024-12-09 Added MiniVLA configs with Qwen2.5 0.5B backbone；All code is made available under the MIT License；pretrained models may inherit restrictions from the underlying base models（Llama Community License）；LIBERO / WidowX
Activity signals
Public counters fetched daily from GitHub, Hugging Face and ModelScope. Numbers only; no judgement. · fetched 2026-10-31
GitHub stars
381
±0 / 30 d
Commits · 90 d
0
last push 592 d ago
HF downloads · 30 d
—
weights not on Hugging Face
Weights last modified
—
ModelScopeno same-name mirror found (searched by repo name)
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 |
|---|---|---|---|
| Other benchmarks | libero_90_success | 61.4 | official · Blog table 'Trained to 95% Action Accuracy', row 'MiniVLA', column 'Libero 90' |
| Other benchmarksMiniVLA + VQ h8 (action chunks) | libero_90_success | 77 | official · Blog table 'Model (Final Action Accuracy) | Libero 90', row '+ VQ h8 (action chunks) (85%)' |
| Other benchmarksMiniVLA + VQ h8 + history=2 | libero_90_success | 82 | official · Blog table 'Model | Libero 90', row '+ VQ h8 + history=2' |
| Other benchmarksMiniVLA + VQ h8 + wrist images | libero_90_success | 82.1 | official · Blog table 'Model | Libero 90', row '+ VQ h8 + wrist images' |
| SimplerEnv · WidowX (Bridge)MiniVLA + VQ h8 (Bridge V2) | put_carrot_on_plate | 44 | official · Blog table 'Model (Final Action Accuracy) | Place carrot | Place spoon | Block stack | Pick eggplant', row 'MiniVLA + VQ h8 (action chunks) (80%)', column 'Place carrot' |
| SimplerEnv · WidowX (Bridge)MiniVLA + VQ h8 (Bridge V2) | put_eggplant_in_yellow_basket | 14 | official · Blog table 'Model (Final Action Accuracy) | Place carrot | Place spoon | Block stack | Pick eggplant', row 'MiniVLA + VQ h8 (action chunks) (80%)', column 'Pick eggplant' |
| SimplerEnv · WidowX (Bridge)MiniVLA + VQ h8 (Bridge V2) | put_spoon_on_towel | 68 | official · Blog table 'Model (Final Action Accuracy) | Place carrot | Place spoon | Block stack | Pick eggplant', row 'MiniVLA + VQ h8 (action chunks) (80%)', column 'Place spoon' |
| SimplerEnv · WidowX (Bridge)MiniVLA + VQ h8 (Bridge V2) | stack_green_block_on_yellow_block | 70 | official · Blog table 'Model (Final Action Accuracy) | Place carrot | Place spoon | Block stack | Pick eggplant', row 'MiniVLA + VQ h8 (action chunks) (80%)', column 'Block stack' |
Deployment recipe
Not reproduced yet · Weights are in the prismatic format and need the openvla-mini runtime rather than transformers remote code; not set up in phase one.
Record history
2026-09-05
Sources fetched
release_date,params_b,license,arch_notes,cross_embodiment,code_url,weights_url
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) | — | — | — | — no latency data | No data |
NVIDIA RTX 5090 ¥2.4/h · Runyun 2.4 / GongJi 2.5 / Paratera Cloud 2.98 (2026) | — | — | — | — no latency data | No data |
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 | — | — | — | — no latency data | No data |
No latency measurements for this model yet · Data comes from literature, maintainer runs, or crowd CLI uploads. None exist so far.
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 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| MiniVLA |
Legend Officially supportedCommunity reproducedFine-tunable in theoryUnsupportedUnknown Vendor-declared target form 0 / 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.