Fern 0.1: a 0.1–0.5T coding model one researcher pretrained in 3.5 months for under $10M
teortaxesTex · x · 2026-10-27
Early preview of Fern 0.1, built on Mirendil-Base (0.1–0.5T), pretrained from scratch by one researcher in 3.5 months for under $10M, with agents writing 99%+ of the code. Targets high-volume AI-builder work (training-data audits, log debugging, agent trace reading), offers 10M-token context (23MB code), 100% single-needle RULER at 10M, and 73.4% SWE-bench Veri… in one agentic attempt via OpenCode. Quoted poster teortaxesTex criticizes the vague "0.1–0.5T" parameter disclosure as unserious compared to models with exact total/active parameter counts.
More from coding & agent
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27
- 'A human approves everything' is a trap: approval fatigue and how to actually design safe agents — Efistoffeles · 2026-10-27