Tsinghua University AIR (2toINF)
X-VLA
880M Commercial use OK 8/8 verified
Facts
- Released
- 2025-10 1 Source
- Parameters
- 0.88 B 1 Source
- Architecture
- Soft-prompted transformer 0 Source
- License
- Apache-2.0 (code and model card) 2 Source
- Commercial
- Commercial use OK 0 Source
- Weights
- huggingface.co ↗ 1 Source
- Code
- github.com ↗ 1 Source
- Paper
- arxiv.org ↗ 0 Source
- Input modalities
- rgb · language · state 0 Source
- Action space
- Cross-embodiment soft prompts, single-arm/dual-arm 0 Source
- Cross-embodiment
- CALVIN / LIBERO / VLABench / RoboTwin2 / BridgeDataV2；Franka / AgiBot G1 / Google Robot / WidowX / AgileX 1 Source
- Pretraining data (hours)
- Unverified 0 Source
- Target forms
- Single arm · Dual arm 0 Source
Architecture · Soft-Prompted Transformer, Florence-2-large vision; 0.9B scale; ICLR 2026
Sources · 2 sources
license,release_date,cross_embodiment,code_url https://github.com/2toinf/X-VLA Official page · Fetched 2026-09-04 · No snapshot stored yet README：Apache License 2.0；arXiv 2510.10274（2025-10）；ICLR 2026；Franka / Agibot-G1 / Google Robot / WidowX / Agilex
params_b,weights_url,license https://huggingface.co/api/models/2toINF/X-VLA-Pt Model hub page · Fetched 2026-09-04 · No snapshot stored yet safetensors 879,738,545；license: apache-2.0；base microsoft/Florence-2-large
Activity signals
Public counters fetched daily from GitHub, Hugging Face and ModelScope. Numbers only; no judgement. · fetched 2026-10-18
GitHub stars
744
+14 / 30 d
Commits · 90 d
0
last push 131 d ago
HF downloads · 30 d
7,247
all-time 129,426 · 14 likes
Weights last modified
2025-11-16
337 d ago
GitHub2toinf/X-VLA · 76 forks · 35 open issues · no GitHub release
ModelScopeno same-name mirror found (searched by repo name)
Hugging Face2toINF/X-VLA-Pt
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 |
|---|---|---|---|
| CALVINX-VLA-0.9B | abc_d | 4.4 avg_len | paper · Table 2, row 'X-VLA (Ours)', Calvin ABC -> D; Table 14 chain success 97.1 / 92.6 / 88.5 / 84.4 / 78.8, Average 4.43 |
| LIBEROX-VLA-0.9B | avg | 98.1 | paper · Table 2, row 'X-VLA (Ours)', LIBERO Avg (also Table 13 'Average') |
| LIBEROX-VLA-0.9B | goal | 97.8 | paper · Table 2, row 'X-VLA (Ours)', LIBERO Goal (also Table 13) |
| LIBEROX-VLA-0.9B | long | 97.6 | paper · Table 2, row 'X-VLA (Ours)', LIBERO Long (also Table 13) |
| LIBEROX-VLA-0.9B PEFT (LoRA adapters, ~9M params) | long | 83.2 | official · README results table, row 'X-VLA-libero-long-peft' |
| LIBEROX-VLA-0.9B | object | 98.6 | paper · Table 2, row 'X-VLA (Ours)', LIBERO Object (also Table 13) |
| LIBEROX-VLA-0.9B | spatial | 98.2 | paper · Table 2, row 'X-VLA (Ours)', LIBERO Spatial (also Table 13) |
| LIBEROX-VLA-0.9B PEFT (LoRA adapters, ~9M params) | spatial | 96.2 | official · README results table, row 'X-VLA-libero-spatial-peft' |
| Other benchmarksX-VLA-0.9B | pdms | 87.3 PDMS | paper · Table 2, row 'X-VLA (Ours)', NAVSIM PDMS; Table 11 NC 97.5 / DAC 96.5 / EP 82.2 / TTC 92.9 / C 100.0 / PDMS 87.3 |
| RoboTwin 2.0X-VLA-0.9B | easy | 70 | paper · Table 2, row 'X-VLA (Ours)', RoboTwin-2.0 Easy (Table 16 'Average' over 50 tasks) |
| RoboTwin 2.0X-VLA-0.9B | hard | 39 | paper · Table 2, row 'X-VLA (Ours)', RoboTwin-2.0 Hard (Table 16 'Average') |
| SimplerEnv · Google RobotX-VLA-0.9B | variant_aggregation | 75.7 | paper · Table 2, row 'X-VLA (Ours)', Simpler VA; Table 12 per task 85.5 / 79.8 / 61.9 / 75.7, Average 75.7 |
| SimplerEnv · Google RobotX-VLA-Google-Robot checkpoint | variant_aggregation | 76.4 | official · README results table, row 'X-VLA-Google-Robot', Google Robot (VA) |
| SimplerEnv · Google RobotX-VLA-0.9B | visual_matching | 80.4 | paper · Table 2, row 'X-VLA (Ours)', Simpler VM; Table 12 per task Coke 98.3 / Near 97.1 / Open 69.5 / Put 56.5, Average 80.4 |
Deployment recipe
No recipe yet. We publish only steps that actually ran on a rented GPU with a measured latency; this model has not been through that yet.
Record history
2026-09-04
Sources fetched
license,release_date,cross_embodiment,code_url · params_b,weights_url,license
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 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| X-VLA |
Legend Officially supportedCommunity reproducedFine-tunable in theoryUnsupportedUnknown Vendor-declared target form 4 / 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.