X Square Robot
WALL-OSS
4.22B 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
- 2025-09 1 Source
- Parameters
- 4.22 B 1 Source
- Architecture
- VLM + action expert 0 Source
- License
- Apache-2.0 (code repository); weights model card has no licence tag 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 · state 0 Source
- Action space
- Dual-arm joints (flow matching / FAST tokens) 0 Source
- Cross-embodiment
- Large-scale multimodal pretraining + real-robot manipulation (README lists no embodiments) 0 Source
- Pretraining data (hours)
- Unverified 0 Source
- Target forms
- Dual arm 0 Source
Architecture · Qwen2.5-VL backbone + MoE action expert (Qwen2_5_VLMoEForAction); FLOW / FAST action heads; 2026-05 released version 0.5 and world model WALL-WM
Sources · 2 sources
release_date,license,arch_notes,code_url https://github.com/X-Square-Robot/wall-x Official page · Fetched 2026-09-04 · No snapshot stored yet README: 2025-09 WALL-OSS; 2026-05 Wall-OSS-0.5 and WALL-WM; Apache-2.0 license; Qwen2_5_VLMoEForAction
params_b,weights_url https://huggingface.co/api/models/x-square-robot/wall-oss-flow Model hub page · Fetched 2026-09-04 · No snapshot stored yet safetensors 4,224,041,072; model card has no license field; base qwen2_5_vl
Activity signals
Public counters fetched daily from GitHub, Hugging Face and ModelScope. Numbers only; no judgement. · fetched 2026-10-31
GitHub stars
1,306
+25 / 30 d
Commits · 90 d
3
last push 44 d ago
HF downloads · 30 d
2,869
all-time 15,299 · 34 likes
Weights last modified
2026-06-02
152 d ago
GitHubX-Square-Robot/wall-x · 107 forks · 53 open issues · no GitHub release
ModelScopeX-Square-Robot/wall-oss-flow · 305 downloads · updated 2026-06-02 · same org as Hugging Face
Hugging Facex-square-robot/wall-oss-flow
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 | Action Planning | 69 | paper · Table 2, row 'WALL-OSS', column 'Action Planning' |
| Other benchmarksWALL-OSS (Co-training) | Letter | 87 | paper · Table 3, row 'Letter', column 'WALL-OSS (Co-training)' |
| Other benchmarksWALL-OSS (Co-training) | Number | 95 | paper · Table 3, row 'Number', column 'WALL-OSS (Co-training)' |
| Other benchmarks | Object Grounding | 91.6 | paper · Table 2, row 'WALL-OSS', column 'Object Grounding' |
| Other benchmarks | Scene Captioning | 87.6 | paper · Table 2, row 'WALL-OSS', column 'Scene Captioning' |
| Real-robot evaluation (self-reported) | Collect-Waste (in-distribution) | 100 | paper · Section 5.2.3 text (Figure 7 bars): 'the pre-trained models—WALL-OSS and pi0—achieve a 100% success rate on the in-distribution (ID) task' |
| Real-robot evaluation (self-reported) | zero-shot pick-and-place, seen-object instructions | 85 % progress | paper · Section 5.2.2 text: 'The model achieves an average task progress of 85% on seen-object instructions' |
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
release_date,license,arch_notes,code_url · params_b,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 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| WALL-OSS |
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.