Live market data · prices move with supplier bids · data date 2026-11-20
grok-4.20-0309-reasoning vs grok-4.20-0309-non-reasoning — which should you pick? Judge by live market data, not marketing: input price $0.198288 / 1M tok vs $0.012558 / 1M tok, 9 vs 7 suppliers, stability — vs — (data date 2026-11-20). grok-4.20-0309-non-reasoning's input price is about 6% of grok-4.20-0309-reasoning's — roughly 15.8× cheaper. Both run on one endpoint — https://321ai.cc/v1 — and switching costs nothing. Failures never billed.
| Dimension | grok-4.20-0309-reasoning | grok-4.20-0309-non-reasoning |
|---|---|---|
| Input price (market floor) | $0.198288 / 1M tok | $0.012558 / 1M tok |
| Suppliers quoting | 9 | 7 |
| Stability | — | — |
| P50 latency | 6521 ms | 3060 ms |
Machine-readable prices: GET https://321ai.cc/api/public/model-prices. Chinese version: https://321ai.cc/compare/grok-4.20-0309-reasoning-vs-grok-4.20-0309-non-reasoning.
Both models live behind one endpoint. Set base_url = https://321ai.cc/v1, then pass either model name as model — zero migration cost, and conditional routes re-pick the best lane within 5 minutes as prices move.
grok-4.20-0309-non-reasoning's input price is about 6% of grok-4.20-0309-reasoning's — roughly 15.8× cheaper.
grok-4.20-0309-reasoning currently has 9 suppliers quoting it — more suppliers means automatic failover when one lane degrades.
Yes. 321AI is fully OpenAI-protocol compatible: set base_url to https://321ai.cc/v1 and switch the model parameter between the two — no code changes.
No. Multiple suppliers bid on every model; this page shows the live market floor, which moves with the market. Per-supplier quotes are on the marketplace page.