frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: JevBench, a reproducible benchmark for typed decision models

https://benchmarkheaven.com/jev-models
26•florianstandhar•8h ago
Hi HN! I built JevBench because Jev kicks ass, and the world deserves to know how the serious open source and fake lookalike projects really perform in comparison.

Jev-class models return bounded choices and probabilities instead of text, and are disruptively faster and cheaper than LLMs, while being similarly intelligent on the text input they operate on.

JevBench allows looking at accuracy, latency and price all at once, in a weighted way - you can even configure the weighting.

A full run asks 534 English decisions. The v1.3 score combines chance-corrected Intelligence, Calibration, Speed and Cost.

Leaderboard right now:

  #1 - Jev            74.4
  #2 - SemIf          73.1
  #3 - djev           73.0
  #4 - Winnow-12B Q8  71.2
  #5 reflex 4B        70.3.
MIT harness, public items, frozen artifacts, scoring code and public per-task outcomes:

https://github.com/fstandhartinger/jevbench

Two no-signup demos:

https://who-is-right.app.mintapis.com

https://is-it-ai-slop.app.mintapis.com

Limitations: English-only; latency from one German server; local/demo latency gets a disclosed ×2 adjustment (+150 ms on my servers) which is an informed assumption; held-out prompts still reach evaluated services; ~1-point gaps can be noise.

Wdyt?

Comments

swyx•23m ago
jev ceo on why he eschewed benchmarking: https://www.latent.space/i/216783460/privacy-benchmarking-an...

Show HN: Training a model to identify AI web content from structure alone

https://arxiv.org/abs/2609.15369
19•jochenmadler•8h ago•5 comments

Show HN: JevBench, a reproducible benchmark for typed decision models

https://benchmarkheaven.com/jev-models
26•florianstandhar•8h ago•1 comments

Show HN: Drop – A rootless Linux sandbox with gVisor support

https://droprun.sh/
141•mixedbit•7h ago•49 comments

Show HN: AI·rete·RAG – a Rete rule engine decides, RAG explains why

https://ai-rete-rag.com/
29•ZaharaHussain•4h ago•2 comments

Show HN: InstinctFlash – High-Performance Serving Runtime for Robotics Models

https://github.com/General-Instinct/InstinctFlash
17•guanming0717•5h ago•2 comments

Show HN: FreeCoffee – Self-hostable donation tool with crypto support

https://github.com/freecoffee-bio/freecoffee.bio
2•jacksun788•2h ago•2 comments

Show HN: Notes on Agentic AI – A text-first guide for practicing engineers

https://github.com/mrsachindixit/agentixit
2•tbaadm•2h ago•0 comments

Show HN: RUUN – Tor Browser with a second, non-Tor mode in the same app

https://gitlab.com/varad_7/ruun
2•opentestudox07•2h ago•0 comments

Show HN: Hack – my Hacker News mobile client now lets you block AI posts

4•busymom0•6h ago•0 comments

Show HN: graf (1000x faster graphify in Rust)

https://github.com/ctxrs/graf
3•ripped_britches•4h ago•1 comments

Show HN: Mini-AGI – Dynamic continual learning model trained on 8GB VRAM

https://github.com/volotat/mini-AGI/
271•volotat•1d ago•72 comments

Show HN: OpenMCP – An open, code-first fork of the Model Context Protocol

https://github.com/enclawed/omcp
2•enclawed•5h ago•0 comments

Show HN: Foremerge – Catch intent conflicts between parallel coding agents

https://github.com/naw103/foremerge
45•naw103•1d ago•15 comments

Show HN: Venya lets AI agents use secrets without seeing them

https://github.com/tabith-llc/venya
2•tabith•7h ago•0 comments

Show HN: A competition for small neural networks that play strategy games

https://tinybrains.dev
109•codetiger•2d ago•40 comments

Show HN: Lossless-memory – a personal AI memory that never summarizes

https://github.com/aru-labs/lossless-memory
64•aru-labs•1d ago•29 comments

Show HN: Radius – A Meetup.com Alternative

https://radius.to/
162•radius89•2d ago•80 comments

Show HN: Matcha – fresh job postings pulled directly from company career pages

https://matchajobs.co/
2•jpercy•9h ago•3 comments

Show HN: Vellum, the best diagram editor you'll ever use

https://vellum.blueprintr.io
21•JoshJamesAnthon•1d ago•15 comments

Show HN: Watch all the AI agents on your machine

https://github.com/markwylde/all-your-agents
5•turblety•10h ago•2 comments

Show HN: An atlas of system designs with interactive architecture diagrams

https://atlas-sysdes.vercel.app/
3•mertkahyaoglu•10h ago•2 comments

Show HN: jevals – replacing LLM judges with typed Jev decisions

https://github.com/openlayer-ai/jevals
44•gbayomi•1d ago•6 comments

Show HN: Sigabrt.dev – cronjob monitor with an SSH TUI

https://sigabrt.dev
81•4815162342•3d ago•35 comments

Show HN: CUA-S1 – A System One Model for Computer Use

https://github.com/trycua/cua
92•frabonacci•3d ago•10 comments

Show HN: Gdocs-me-up: a high-fidelity Google Docs exporter

https://github.com/behdad/gdocs-me-up
20•behdad•1d ago•8 comments

Show HN: Share your AI Setup, Learn from others

https://mysetup.ai/
249•steveybrown•5d ago•139 comments

Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash

https://cactuscompute.com/needle
236•HenryNdubuaku•4d ago•92 comments

Show HN: Snapdrop: Instantly share files between devices. No setup, no signup

https://snapdrop.me
111•Capira•5d ago•51 comments

Show HN: Jev Powered Obsidian Search

https://github.com/Emlembow/jev-graph-search
4•MikeLembo•21h ago•1 comments

Show HN: Lightspeed, build real time apps the Laravel way (+asteroids demo)

17•jv22222•1d ago•5 comments