Tiny model frontier mapped: GLiNER 2.5 leads <350M, Kev family tops two brackets
multimodalart · x · 2026-09-22
multimodalart maps out the "tiny frontier" of small models: GLiNER 2.5 leads the <350M bracket, the Kev family tops both 350M-700M and 3B-10B, Decider 2B sits in the sub-3B sweet spot, and Jevfire's Qwen3.8 27B single-pass implementation is the best open-weights jev-like model.
Follow-up notes that jev-style models do particularly well on graduate-level knowledge questions, Python function choosing, tool calling and model routing. The latency-performance Pareto frontier also includes openvons' Qwen3-4B as a super-fast option and the now-classic diffusiongemma jev-like implementation.
More from Fun
- Developer uses high-speed AI agents to clean up old GitHub repos: "make the speed force count" — MickeySteamboat · 2026-09-22
- Norwegian Consumer Council's 'enshittification' explainer video hits 2.5M views — LuizaJarovsky · 2026-09-22
- AI generation fails the coffee but nails the crystal glass's best angle — misovalko · 2026-09-22
- A type design style becomes the new em-dash: an LLM signature people now recognize on sight — RileyRalmuto · 2026-09-22
- Claude Field trends with music boxes, a rain machine and 'The Rim Remembers The Hand' — RileyRalmuto · 2026-09-22
- Pelican SVG test puts unreleased GPT-6 Astra head-to-head with Anthropic's Mythos — PrisonOfH0pe · 2026-09-22