ignis · model provider pricing
Per 1M tokens. CNY with USD at 1 USD = 7.10 CNY.
| Provider | Model | Input miss | Cached input | Output | Source |
|---|---|---|---|---|---|
| DeepSeek | deepseek-v4-flash | ¥1.00 · $0.14 | ¥0.02 · $0.003 | ¥2.00 · $0.28 | api-docs.deepseek.com |
| Moonshot | kimi-2.6 | ¥6.50 · $0.92 | ¥1.10 · $0.15 | ¥27.00 · $3.80 | platform.kimi.com |
| MiniMax | MiniMax-M3 | ¥2.10 · $0.30 | ¥0.42 · $0.06 | ¥8.40 · $1.18 | platform.minimaxi.com |
| Zhipu | glm-5.1 | ¥6.96 · $0.98 | ¥1.85 · $0.26 | ¥21.87 · $3.08 | openrouter.ai/z-ai/glm-5.1 |
| DeepSeek | deepseek-v4-pro | ¥12.35 · $1.74 | ¥0.10 · $0.0145 | ¥24.71 · $3.48 | api-docs.deepseek.com |
- kimi-2.6 is billed via a flat-fee Coding Plan in practice; rates above are the per-token equivalent.
- MiniMax-M3 row uses the ≤512k-context promotional rate.
- glm-5.1 was run via the Ark Coding Plan (flat-fee); rates above are the per-token equivalent. For glm-5.1, Ark's
/api/coding/v3endpoint returned no cached-input (all input billed at the miss rate); deepseek-v4-pro on the same endpoint does cache (~76%). - deepseek-v4-pro shown at standard rates — the 75%-off promo ended 2026-05-31.