Terse is a family of ternary-weight language models — every internal weight is −1, 0 or +1. Own architecture, trained from scratch, in the open, under Apache 2.0. The pitch isn't a leaderboard score — it's the most capability per megabyte and per joule: open, private, and running on the device in your hand.
Five ternary models on a single ladder — every one ownable, offline, and built on {−1, 0, +1}. Terse Micro is the foundation: the architecture proven in Micro is the base every larger tier inherits. Only Micro exists today — the rest are planned, described here by the hardware they'll run on, not by specs we haven't earned yet.
~423M params (~320M active). Runs on any phone or CPU. Ships as a ~182 MB file.
Laptop-class. Built on the Micro base — the next step up the ladder.
Runs interactively on an ordinary 16 GB-RAM laptop — no discrete GPU.
Multimodal. Runs on a gaming laptop — 16 GB RAM plus a 4 GB GPU.
Frontier-scale on a single server node. Fully multimodal.
Only Terse Micro exists today. Mini, Medium-Lite, Medium and Pro are planned — described by target hardware and modality; exact sizes and results come when they're trained.
A ~423M-parameter mixture-of-experts model, trained from scratch in pure PyTorch on 8 billion tokens, then chat-tuned (SFT) and identity-aligned (ORPO). It's a proof-of-concept and the foundation of the family — the model that validates the architecture the larger tiers inherit. The headline isn't a benchmark; it's size, efficiency, portability and openness.
Ternary weights carry ~1.6 bits of information each; the model ships as one TQ2_0 GGUF (~182 MB, Q6_K embeddings).
Runs on a plain CPU via the terse-arch llama.cpp fork — no GPU required.
FineWeb-Edu · ~19 tokens/param — Chinchilla-reasonable for its size.
The pretrained language model — 8B tokens of FineWeb-Edu. Fluent English continuation; GPT-2-medium-class by design at this data budget.
Chat model — 3-epoch supervised fine-tune on a 44.5k-conversation corpus. Learns the chat template and turn structure.
Identity-aligned with ORPO on top of the SFT model. Preference alignment measurably shifts it toward its own charter.
Terse has a deliberate identity, tuned in during alignment: English-first (bilingual is the goal for later tiers), built to democratize AI — capable, private, local, free of paywalls. Developed in 2026 by Michelangelo Romero Chisco.
It's honest about its limits — it owns being a small model rather than bluffing, stays neutral on contested topics, and keeps your data on your machine. Playful when the moment's casual, precise when there's real work on the table. No stereotypes, no grandstanding.
The single-file model, code, and inference fork are all public.