Last updated: 2026-09-20 · Rates verified against provider listings on this date
Jev Pricing: What It Actually Costs (2026)
Jev's reference price is $0.042 per 1 million input tokens — about $42 per billion — and output is free on every provider, because the model returns compact typed decisions instead of generated text. This page tracks every provider's rate, translates them into real per-decision costs, and lets you compute your own bill.
Official pricing
| Item | Rate | Notes |
|---|---|---|
| Input tokens | $0.042 / 1M | The only thing that is billed |
| Output tokens | $0 | “Too cheap to meter” — answers are typed decisions, not text |
| Context window | 64,000 | 32,000 on Cloudflare; state must fit |
| Typical decision | ~$0.0001 | Official benchmark case: $0.000081; vendor eval: $0.0004/case |
TypeSafe's $0.042/1M input rate is roughly 238x lower than “Claude Fable 5.1” and about 1/48th of GPT-5.6 Terra's $2.00/1M — and those models also bill $2–$12 per 1M output tokens, which Jev never charges.
Provider comparison
Jev is served by four public providers plus TypeSafe's own console. Same model, different context limits:
| Provider | Model ID | Input / 1M | Output / 1M | Context | Waitlist? |
|---|---|---|---|---|---|
| TypeSafe console | jev-latest | $0.042 | $0 | 64K | Yes (early access) |
| Vercel AI Gateway | typesafe-ai/jev | $0.042 | $0 | 64K | No |
| OpenRouter | typesafe/jev-1.13 | $0.042 | $0 | 64K | No |
| Cloudflare Workers AI | typesafe/jev | $0.042 | $0 | 32K | No |
| Vivgrid | jev | $0.042 | $0 | 64K | No |
Cost calculator
Estimate your monthly bill. Jev charges input only; chat models bill both sides (assumes ~60 output tokens per decision for LLMs):
Rates from provider listings, Sep 2026. Real per-decision cost depends on your state length — keep states lean.
Real-world cost reports
- 1,018 research papers classified for $0.08 total, 256ms median latency per paper (community report, Sep 2026).
- Production dispatch workflow: −71% cost and −90% wall time after switching the decision step to Jev.
- Order-flow benchmark: $0.039 per 1,000 decisions, 0.49s median on live data (independent test).
- Video debate scoring (jevmeter): every sentence scored for about $0.05 total.
Is the price sustainable?
TypeSafe is unusually candid here: it cannot prove the pricing isn't subsidized and says the long term will have to demonstrate sustainability — though it expects prices to fall rather than rise. Treat the current rate as a launch figure. Our news tracker watches provider listings and will flag any change automatically.
What drives your real cost
- State length dominates. Cost scales with input tokens, so a 300-token lean state costs ~2.7x less than an 800-token padded one.
- Adding questions is nearly free. Questions ride along in the same request; speculative fan-out barely moves the bill.
- Clean states beat cheap models. Stripping HTML boilerplate and metadata reduces tokens and improves accuracy — the double win TypeSafe recommends in its query-design guidance.