Jev (TypeSafe AI) API 定价
面向技术团队Jev (TypeSafe AI) 按使用量而非按月定价；每种选项的费率见下表。 已于 2026年11月10日 对照官方定价页面核对。
一种决策模型，而非聊天机器人：你发送一个结构化问题，然后返回一个带类型的选项、分数或带有概率的是/否，你的软件可以据此采取行动。早期访问且仅限开发者——它不聊天也不生成文本，没有最终用户应用。
Jev is a decision model, not a chatbot: your software asks a structured question and gets back a typed choice, score or yes/no with a probability it can act on. For a developer running high-volume automations it can replace a general LLM call on simple judgement steps at a fraction of the cost and latency. It is early access, developer-only, and not something a small business can use without an engineer.
套餐和价格
| 套餐 | 价格 | 包含内容 |
|---|---|---|
| Early access API我们的推荐 | 按使用量$42 per billion input tokens = $0.042 per 1M input tokens; output tokens free. Early-access pricing, subject to change at GA. | Pay per use
|
选择这一档：Early access API
It is the only way to use Jev today. Start with one narrow, well-labelled decision such as ticket routing or lead scoring, and compare it against the LLM step it replaces.
优点
- Returns typed decisions with calibrated probabilities instead of free text
- Very low per-token price, with output tokens free
- Much faster than a general LLM on classification-style steps
- Available through Vercel AI Gateway, Cloudflare Workers AI and LangChain
注意事项
- Early access: pricing may change at general availability
- Cannot chat, write or summarise; it only decides
- Independent tests measured far smaller speed and cost gains than TypeSafe claims
- A reported prompt-injection test shifted its verdict on a dangerous command
- No end-user app; needs a developer to integrate
以下情况可跳过
- You do not have a developer building automations
- You need text generated, not a decision made
- The decision is safety-critical and Jev would be the only guardrail
- You need stable pricing and a general-availability product
最适合
Jev (TypeSafe AI) 定价问题
- Jev (TypeSafe AI) 在 2026 年要多少钱？
- Jev (TypeSafe AI) 按使用量而非按月定价；每种选项的费率见下表。 已于 2026年11月10日 对照官方定价页面核对。
- Is Jev a chatbot like ChatGPT?
- No. Jev does not generate text. It answers structured questions with a choice, a score or a yes/no, plus a probability, so another program can act on the result.
- Who makes Jev?
- TypeSafe AI, which launched Jev in limited early access in September 2026.
- What is Jev good for?
- High-volume judgement steps inside automations: lead scoring, support-ticket routing and urgency, content moderation, ranking, allow/block/review checks for AI agents, and deciding when a task needs a more expensive model.
- Can I rely on Jev as a security guardrail?
- Not on its own. A test reported by VentureBeat showed that injected text posing as a tool result could lower Jev's confidence that a dangerous command should be blocked. Keep a hard rule or a human check in front of anything irreversible.
- Are TypeSafe's speed and cost claims independent?
- No, they are vendor-reported. Latent.space's own comparison found real but much smaller gains than TypeSafe advertises.
由 Dapols team 于 2026年9月22日 审阅。 报告价格变动
Jev (TypeSafe AI) 适合你的业务吗？
回答几个问题，看看 Jev (TypeSafe AI) 是否适合你的工具组合和预算。