The Qwen3 routing brain auto-selects the best model by task capability. Pay only for what you use — cache hits cost even less.
| model | input $ | output $ | cache hit | notes |
|---|---|---|---|---|
| mesh-free | $0 | $0 | — | daily free quota |
| mesh-pro | $0.30 | $0.60 | $0.005 | intelligent routing workhorse |
| mesh-max | $0.90 | $2.00 | $0.005 | single flagship model |
| mesh-research | $3.00 | $8.00 | — | deep research MoA |
| node tier | device | credit multiplier | models |
|---|---|---|---|
| Tier 0 | Phone | ×0.3 | 0.5-2B |
| Tier 1 | Light PC | ×1 | Qwen3-8B |
| Tier 2 | Mid-range GPU | ×3 | Qwen3-14B |
| Tier 3 | Powerful GPU / large-memory personal rig | ×10 | GLM-5 / 32B / V4-Pro (DGX Spark, M5 Ultra) |
| Tier 4 | Compute center (enterprise) | ×30 | H100/A100 running GLM-5.2 / V4-Pro |
Credits are spendable on mesh-pro/max/research within the platform. Credits are not withdrawable for cash — platform use only.
500 free credits on sign-up, no credit card required