Ling 3.1 Flash
工具调用推理缓存
官方价格
¥0.5 / ¥1.5
Ling official platform · 输入 / 输出 · 每百万 token · Limited-time 75% discount; cache reads ¥0.10; checked 2026-10-16
最低输入价
$0.045
Novita
上下文
1M
基础价格
1最新价格检查于 Nov 18, 2026。
Ling 3.1 Flash · 已公布价格
| Billing category | Input / 1M | Output / 1M | Cache read / 1M |
|---|---|---|---|
| Official platform promotion | ¥0.50 | ¥1.50 | ¥0.10 |
| Official platform original | ¥2.00 | ¥6.00 | ¥0.40 |
| OpenRouter promotion (third party) | $0.075 | $0.225 | $0.015 |
| OpenRouter original (third party) | $0.30 | $0.90 | $0.06 |
Ling 3.1 Flash · 权重许可证
| 模型权重 | 协议 |
|---|---|
| 开放权重 | MIT |
缓存与处理模式
The checked pricing page lists a limited-time 75% discount: ¥0.50 per million input tokens, ¥1.50 per million output tokens, and ¥0.10 per million cache-read tokens.
[2]
上下文与限制
The page says original prices will gradually resume after the promotional period; it does not provide a fixed end date in the checked text.
[2]It also says the official service currently exposes 256K context.
[3]Ling’s official model documentation lists roughly 560 billion total parameters and 25 billion active parameters for Ling-3.1-flash, with support for up to one million context tokens.
[3]
关于 Ling 3.1 Flash
The official Ling platform lists a limited-time 75% discount at ¥0.50 input, ¥1.50 output, and ¥0.10 cache read per million tokens. Original rates are ¥2.00, ¥6.00, and ¥0.40.
No. The official document places its USD prices under Third-Party Platforms, including OpenRouter and ZenMux. Direct-platform prices are in CNY. Check actual provider quotes before purchasing.
The checked pricing page calls it a limited-time promotion and says original prices will gradually resume. It does not give a fixed end date in the checked text.
Ling lists roughly 560B total parameters, 25B active parameters, and up to 1M context capability. The official service currently exposes 256K context; other endpoints can have different limits. Use the official documentation links for weights and license details.
Use the promotional rate for a scenario only when that is the schedule applicable to the account and billing period. Retain an original-rate scenario for longer-term planning because the official page says the promotion is temporary and original prices will gradually resume. The checked text supplies no fixed end date, so do not schedule a fictional expiry in the calculation. For a consistent workload with ten million uncached input tokens and two million output tokens, the official CNY examples are ¥8 at promotional rates and ¥32 at original rates. Those figures exclude unknown account-specific charges and do not predict future usage. Keep the tariff source date with the forecast. If rates change later, produce a new estimate instead of silently rewriting the assumptions behind a historical report.
The pricing document separates the direct official platform, quoted in CNY, from Third-Party Platforms quoted in USD. They are different published schedules, not exchange-rate conversions supplied by this guide. The OpenRouter reference also tells readers to verify actual platform pricing. A comparison therefore needs the serving platform, current provider route, currency, applicable promotion, cache policy, and source date. Converting currencies, if required, needs an additional exchange-rate source and does not make the service conditions identical. Keep the original CNY tariff visible when reporting official direct-platform pricing. Keep the USD offer associated with its third-party platform rather than describing it as a direct Ling price. An API route’s context limit should also be checked separately: the official documentation currently exposes 256K despite the model’s up-to-1M capability.
来源与注脚
- [1]来源核验日期:
- [2]
- [3]
输入、输出和缓存读取的单位是每百万 token 的美元价格。速度是每秒输出 token 数。可用率是来源报告的在线比例。