Jev 1.13 is the earlier release of TypeSafe's decision model: you describe a situation and it hands back a choice from options you set, a score, or a yes or no with how sure it is, never a written reply. It takes in about 260 pages at once. Keep it when a routing or policy check you already run was tuned on this version; for anything new, Jev 1.14 is the current release of the same model.
Kev 4B
by JaredpalmerReleased 25 September 2026
Kev 4B is a decision model you can download and run yourself: you send a situation as text plus questions of a fixed type (yes or no, pick one of several, or a score) and get one calibrated probability per question, with no written text. Use it for sorting requests, labelling content and rule checks on text of up to 8 pages. It is the recommended size in the Kev family and comes under the Apache-2.0 licence.
Where your request runs
One company runs this model. OpenDunes sends every request to it. DA per 1M tokens: ↑ what you send.
You pay the price of the provider that served your request. Prices read 02:48 · speed over the last 30 minutes · uptime over 3 days.
What it costs
You pay for what each request uses. No subscription, no minimum.
One case, the ordinary use
Every rate
| Meter | ↑ you send | ↓ comes back |
|---|---|---|
| Text tokens | 13,98 DA |
per 1M tokens
Everyday jobs, priced
- When you pay
- After the answer, from your balance. Nothing is held.
- Why it varies
- The length of the text.
- After each one
- The exact charge shows on the request.
Call it in code
One HTTP call, with your questions named as you like.
import os, requests
BASE = "https://opendunes.com"
KEY = os.environ["OPENDUNES_API_KEY"]
H = {"Authorization": f"Bearer {KEY}"}
resp = requests.post(f"{BASE}/alpha/decisions", headers=H, json={
"model": "jaredpalmer/kev-4b",
"state": "The courier came two days late and the box was open.",
"questions": {
"is_complaint": {
"type": "noul",
"instructions": "Is it a complaint?",
"criteria": {
"true": "The customer is unhappy with the service.",
"false": "The customer is asking a question or saying thanks.",
},
},
},
})
print(resp.json())Runs this request once on your balance.
Limits
The hard numbers — what it can take in, what it can give back, and where it stops.
- Context window
- 8 192 tokens
- Max output
- 7 372 tokens
- Reads at once
- ≈ 8,1 pages
- Longest reply
- ≈ 6 000 words
- Tokenizer
- Other
- Released
- 25 Sep 2026
Questions about Kev 4B
How much does Kev 4B cost in dinars?
About 0,014 DA per case. ↑ 13,98 DA per 1M tokens.
Is Kev 4B free? Can I try it?
No. You pay per use from your balance in dinars, with no subscription. The chat has free models for free messages every day.
What is Kev 4B best for?
Kev 4B is a decision model you can download and run yourself: you send a situation as text plus questions of a fixed type (yes or no, pick one of several, or a score) and get one calibrated probability per question, with no written text. Use it for sorting requests, labelling content and rule checks on text of up to 8 pages. It is the recommended size in the Kev family and comes under the Apache-2.0 licence.
How do I pay for Kev 4B without an international card?
Add credit in dinars with your Edahabia or CIB card. Every model, Kev 4B included, is paid from that balance. No foreign card, no conversion.
Does Kev 4B work in Arabic, French or Darja?
The model card tags English only. Arabic, French and Darja are not named.
What can it answer?
Yes or no, one choice from your list, or a score on your scale. Each answer comes with its probability.
Who makes Kev 4B, and when was it released?
Jaredpalmer. Released 25 September 2026.
Similar models, priced in dinars
See all modelsJev 1.14 reads a description of a situation and makes a quick decision instead of writing a reply: it picks from options you define, gives a score, or answers yes or no with a probability. It never writes explanations or code, and it reads about 260 pages at once. Use it for routing requests, sorting them into groups and policy checks. Jev 1.13 is the earlier version of the same model.
Decider V1.2 does not write text. You send content as text, a table of data or a picture, plus named questions of three kinds: yes or no, a choice among as many as 255 options, or a score on an ordered scale. It returns the probability of every answer, so your own system sets its thresholds for sorting, routing or moderation. It is the larger model of the pair and reads about 262 pages; Decider Flash V1.2 is the smaller, lower-cost version.
GPT-6 Luna Decisions does not write text. You send content as text, a table of data or a picture, plus up to 200 named questions of three kinds: yes or no, pick among options, or grade against an ordered scale. It returns a probability for every answer, with the likeliest option or expected score, so your own system can set its thresholds for routing, moderation or marking.

