Updated
Desk
Access
Public listings for calling Jev, as of the stories on this desk. Price, context window, and quota are whatever that host published. On September 30, 2026 we sent one POST to api.typesafe.ai/v1/systemone with model jev-latest. The state was "Help! My payouts have been failing for 3 days." The response was HTTP 200 and named model jev-1.13.0. The Noul is_urgent returned 0.95. The Choice department selected billing, with probabilities 0.88 for billing, 0.12 for technical, and 0.0 for sales, and confidence 0.82. The Score frustration returned 1.04. Its legend is 0 Calm, 1 Frustrated, 2 Very angry, with probabilities 0.0, 0.96, and 0.04, and confidence 0.94. Usage was 426 input tokens and 73 output tokens. Vercel, Cloudflare, Netlify, OpenRouter, Venice, MotherDuck, and LiteLLM were not called. Community servers that copy the same HTTP shape sit on repos. djev-run and rev are two of those servers. Rizzo Flow, Jevre, and jev_local are three more. They serve other weights. Jevflake is a dbt package that calls TypeSafe's API from Snowflake SQL. duckdb-jev does the same from DuckDB. prompt_jev() is MotherDuck's hosted SQL function on paid plans. jsort calls the same API, or OpenRouter, once per pair. For how a listing gets on the digest, see method.
TypeSafe API
POST /v1/systemone with a TypeSafe key. Date headers 08:49:11 GMT to 08:49:40 GMT October 26, POSTs with jev-latest and jev-preview named jev-1.14.0. Opened that hour, the models page says both aliases point at jev-1.14.0, and that there is no preview build. Opened October 15, 2026, the models page lists Jev 1.14 and Jev 1.13 at $0.042 per million input tokens, free output, 64k per request and 32k per question. On that reading jev-preview pointed at jev-1.14.0 and jev-latest still pointed at jev-1.13.0. TypeSafe-Danger-Zone-Long-Context: true raises both caps to 262k. Images need TypeSafe-Danger-Zone-Images: true. Date header 00:28:46 GMT October 16, GET /v1/models returned HTTP 200, a POST with jev-latest named jev-1.13.0, and a POST with jev-preview named jev-1.14.0. Date header 14:57:05 GMT, GET /v1/models listed those two aliases, a POST with jev-latest named jev-1.13.0, and a POST with jev-preview named jev-1.14.0. Date header 11:04:52 GMT, GET /v1/models listed those two aliases, a POST with jev-latest named jev-1.13.0, and a POST with jev-preview named jev-1.14.0. An 850 by 850 PNG in state used 2,238 input tokens on jev-1.14.0. Date headers 06:16:20 GMT to 06:16:44 GMT, 32 POSTs named jev-1.14.0 on jev-preview and jev-1.13.0 on jev-latest. Three Chinese payouts tickets on jev-1.14.0 scored frustration mean 1.30 at mean confidence 0.54. Date header 01:22:13 GMT, a POST with model jev-preview returned jev-1.14.0. A POST with model jev-latest returned jev-1.13.0. Date headers 03:03:53 GMT to 03:03:56 GMT, six POSTs with jev-preview named jev-1.14.0. Four payouts tickets in English, Chinese, Spanish, and Japanese came back at language confidence 1.0, all billing. Chinese frustration scored 1.37 at confidence 0.44. Opened October 1, 2026, the models page lists jev-1.13.0 at $0.042 per million input tokens, free output, 100K tokens per second and 40 requests per second, 64k for the whole request and 32k for state plus the longest question. The page says those limits can change without notice. Opened October 7, 2026, the same page lists 80 requests per second. The price, the 100K token rate, and the 64k and 32k caps on that row were unchanged. TypeSafe posted on September 20 that there is no waitlist and that new accounts start with $5 of credit. On September 22 it paused new signups. On September 24 it said signups were still closed. On September 27 at 22:30 UTC it posted that capacity had increased and signups were open, and that new signups no longer get free credits. On September 30, 2026 we sent one POST to that path with model jev-latest. The state was the docs' payouts sentence, and the body asked is_urgent, department, and frustration together. The response was HTTP 200 and named jev-1.13.0. is_urgent, a Noul, returned 0.95. department, a Choice, selected billing, with probabilities 0.88 for billing, 0.12 for technical, and 0.0 for sales, and confidence 0.82. frustration, a Score, returned 1.04. Its legend is 0 Calm, 1 Frustrated, 2 Very angry, with probabilities 0.0, 0.96, and 0.04, and confidence 0.94. Usage was 426 input tokens and 73 output tokens. On October 8, 2026 the same signed-in console still showed available credits of $0.00, no payment method, and no billing history. Usage for the last 7 days was $0.00, 0 tokens, and 0 requests. On October 6, 2026, at 17:05 UTC, a new console account showed available credits of $0.00, no payment method, and no billing history. We did not add a card or send a request from that account.
Vercel AI Gateway
typesafe-ai/jev on the Gateway, with an experimental evaluate method in the AI SDK. Listed September 16. The October 2026 Production Index, published October 15, prints Jev at 35.3% of paid teams in its first two weeks, against Sonnet 5.5 at 18.1%. The September 18 post still prints nearly 13% in the first 24 hours. The model card lists input and output as Free and prints promotional pricing ending September 25, 2026. Opened again on September 26, it still listed Free for both, on TypeSafe AI and DigitalOcean, and it still printed that end date. The form router template calls typesafe-ai/jev through experimental_evaluate and keeps the destination at confidence 0.95, with gpt-5.6-luna-fast on the rest. A Vercel deploy uses an OIDC token. Gateway quotas are not the same as a TypeSafe console key.
Cloudflare Workers AI
Rita Kozlov posted that Jev was live on Cloudflare AI Gateway. Workers AI serves typesafe/jev via env.AI.run. The model card lists a 32,000-token window. Opened October 19, 2026, typesafe/jev-1.14.0 is a second card: $0.042 per million input tokens, output $0.00, 32,000-token window, Box and images marked beta. HTTP 200. We did not call Workers AI. Opened October 8, 2026, the older typesafe/jev card lists the same $0.042 and 32,000-token window. On October 1 Cloudflare posted two other models on the same host, @cf/cloudflare/clef at $0.24 per million input tokens and @cf/cloudflare/clef-flash at $0.09. Those cards are not doors onto TypeSafe's Jev.
Netlify AI Gateway
A September 17 changelog: install @typesafe-ai/sdk, construct TypeSafeClient inside Functions, bill against Netlify credits. Sample routes a contact form between sales, support, and spam.
OpenRouter
Opened October 16, 2026, OpenRouter listed typesafe/jev-1.14 at $0.042 per million input tokens, output free, a 262K context, and TypeSafe P50 0.19 s. The ~typesafe/jev-latest page already linked Jev 1.14. GET /api/v1/models still omitted that id. GET of that id still returned HTTP 404. This desk did not call OpenRouter. Reopened at 14:57 UTC October 15, the HTML page still returned HTTP 200. SosyalDijital's 29 of 30 used recorded fields typesafe/jev-1.13-20260917 and typesafe/jev-1.14-20261014. The call is POST /api/alpha/decisions, not chat completions. On September 24 Diogo Almeida posted a chart of requests in the 1K to 10K token band. The list under it puts Jev 1.13 first, at 242 million requests and 15.1%. The date labels run from Mar 30 to Sep 14. On September 25 OpenRouter posted typesafe/jev-router. The thread says that router solved 237 of 423 tasks and Auto Router solved 130 of 423. The model page we opened documents the chat call and does not reprint those counts. On September 28 Christoph Görn logged sixteen prompts through typesafe/jev-router. The answered_by field names gpt-6-luna five times, deepseek-v4.1-flash three times, and kimi-k3 once, with 7 of 16 calls in error. We did not call the router, and we did not pull a rankings export. Arena's October 7 thread is on the Jev Router page: +3.6% net and a median $0.155 a task. The model page FAQ, opened October 8, says the router has no price of its own and bills the model that serves the request.
Venice
Beta Decisions API on September 18: POST /api/v1/decisions, model jev-latest, same $0.042 rate, 64k / 32k split. Same Venice key as chat and image. Venice says it keeps no copy of the state. The handler is also at /systemone for a TypeSafe SDK base-URL swap. API only; no web-app surface.
MotherDuck
prompt_jev() on paid MotherDuck plans. A SQL statement sends a text column and a choice, score, or yes/no question. The published AG News script reads result.choice and result.confidence. The 100,000-row table is vendor-reported: 89% accuracy, 2,484 rows/s, $0.50, 40 seconds.
LiteLLM
Self-hosted proxy in v1.103.0-rc. TYPESAFE_API_KEY lives on the proxy; clients send a LiteLLM virtual key to POST /typesafe/v1/systemone. Kerry Lu and Mateo Wang's October 15, 2026 post adds day-0 Jev 1.14 images, box, and long context on /v1/systemone and /v1/decisions, and forwards TypeSafe-Danger-Zone- headers to TypeSafe only. Spend is logged as typesafe/jev-1.13.0 even when the request asks for jev-latest. A typesafe guardrail can drop older tool results below a 0.2 relevance score before the generator runs. This desk did not run the proxy.
Databricks
Databricks posted October 16, 2026, 19:17 UTC, that TypeSafe's Jev-1.14 is available through ai_classify and ai_decide on the Lakehouse, governed by Unity Gateway. The September 30 ai_decide blog does not name the checkpoint. This desk did not call Databricks.