Helix 02Figure source ↗Helix 02 vs SmolVLA 450M
At a Glance
| Compare |  Helix 02Figure |  SmolVLA 450MHugging Face LeRobot |
|---|---|---|
| Pricing and Limits | ||
| Context windowMaximum documented tokens | Not reported | Not reported |
| Model facts checked | Aug 29, 2026View model evidence → | Aug 29, 2026View model evidence → |
Available Benchmarks
No Protocol-Matched Benchmark Yet.Results appear here only when both models share the same benchmark version, metric, evaluation protocol, and evidence class.
Helix 02Figure source ↗
SmolVLA 450MHugging Face source ↗Side-by-Side Facts
| Field | Helix 02 | SmolVLA 450M |
|---|---|---|
| Developer | Figure | Hugging Face LeRobot |
| Family | Helix | SmolVLA |
| Model | Helix 02 | SmolVLA 450M |
| Version | 02 | 450M |
| Lifecycle | active | active |
| Released | 2026-01-05 | 2025-06-03 |
| Knowledge cutoff | Unknown | Unknown |
| Input modalities | Text, Image, Robot state | Text, Image, Robot state |
| Output modalities | Robot action | Robot action |
| Context window | Unknown | Unknown |
| Total parameters | Unknown | 450M |
| Active parameters | Unknown | Unknown |
| License | Unknown | apache-2.0 |
| Open weights | No | Yes |
| API available | No | No |
| Self-hostable | No | Yes |
| Provider access | Unknown | Unknown |
| Capabilities | dexterous-manipulation, long-horizon-control, tactile-control, whole-body-control | asynchronous-inference, fine-tuning, low-cost-hardware, manipulation |
| Robotics model type | Vision-language-action model | Vision-language-action model |
| Action representation | Full-body joint targets | Continuous action chunks from a flow-matching action expert |
| Control architecture | Semantic reasoning, visuomotor policy, and kHz whole-body controller | SmolVLM2 backbone with flow-matching action expert |
| Inference location | On device | On device |
| Native control rate (Hz) | Unknown | Unknown |
| Supported embodiments | Figure 03 | SO-100, SO-101, LeKiwi, LIBERO Franka |
| Training data | Figure reports more than 1,000 hours of human motion data plus sim-to-real reinforcement learning for its whole-body controller. | Compatibly licensed LeRobot community datasets totaling fewer than 30,000 episodes in the cited release. |
Helix 02 Capabilities
Model typeVision-language-action model
InferenceOn device
Action representationFull-body joint targets
Supported embodiments1
Canonical IDfigure/helix-02
SmolVLA 450M Capabilities
Model typeVision-language-action model
InferenceOn device
Action representationContinuous action chunks from a flow-matching action expert
Supported embodiments4
Canonical IDlerobot/smolvla_base
Related Comparisons
All robotics comparisons →| A | Pair | B | Context |
|---|---|---|---|
 MolmoAct 2Ai2 | vs | Helix 02Figure | cross-developer peersrobotics |
Helix 02Figure | vs |  Gemini Robotics 2Google DeepMind | cross-developer peersrobotics |
Helix 02Figure | vs |  GR00T N1.7 3BNVIDIA | cross-developer peersrobotics |
Helix 02Figure | vs |  OpenVLA 7BOpenVLA Research Team | cross-developer peersrobotics |
Helix 02Figure | vs |  pi 0.7Physical Intelligence | cross-developer peersrobotics |
Primary Evidence
Sources and Freshness
Questions
Helix 02 vs SmolVLA 450M FAQs
Is Helix 02 or SmolVLA 450M better for coding?+
This comparison does not currently contain a protocol-matched coding benchmark for both Helix 02 and SmolVLA 450M, so Model Markets cannot name a coding leader from pricing, context size, or capability labels alone.
Which is cheaper, Helix 02 or SmolVLA 450M?+
Neither model has a directly sourced input price in this comparison. Neither model has a directly sourced output price in this comparison.
Which has a larger context window, Helix 02 or SmolVLA 450M?+
Neither model has a sourced context window in this comparison, so a larger limit cannot be determined.
Which performs better in benchmarks, Helix 02 or SmolVLA 450M?+
There is no overall benchmark winner: At least two independently verified, protocol-matched benchmarks are required for an overall winner.
Can Helix 02 or SmolVLA 450M be self-hosted?+
SmolVLA 450M is documented as self-hostable; Helix 02 is documented as not self-hostable. Helix 02 is not marked open weight; SmolVLA 450M is open weight.
Can Helix 02 and SmolVLA 450M understand images?+
Helix 02 is documented with image input; SmolVLA 450M is documented with image input. This reflects supported input modalities, not vision quality.
Which can generate longer answers, Helix 02 or SmolVLA 450M?+
Neither model has a sourced maximum output in this comparison, so a larger limit cannot be determined.
Do Helix 02 and SmolVLA 450M support reasoning and tool use?+
Helix 02: image input. SmolVLA 450M: image input. Feature support does not establish relative quality.
Which is available from more inference providers, Helix 02 or SmolVLA 450M?+
Helix 02 has 0 sourced provider routes; SmolVLA 450M has 0, a tie.
Which offers better value, Helix 02 or SmolVLA 450M?+
There is no universal value winner. Compare the input and output prices above with the matched benchmark result for your workload: cheaper tokens can be offset by different quality, token usage, latency, or provider availability.