Mirendil Launches Fern Its
A first model from an obscure lab is worth a raised eyebrow, and the 10M-token window is clearly the headline hook. But the benchmark choice is telling: a perfect score on RULER's single-needle test is the easiest retrieval task in the suite — one key hidden in a haystack — and says little about reasoning across 10M tokens, where most long-context models collapse.… more
The candid admission that Fern 0.1 trails frontier models on agentic coding confirms this is a niche play: retrieval over massive codebases and research corpora, not general competence. In our view, the open question is whether extreme context genuinely pays off in agentic workflows or remains a demo-friendly party trick.
Mirendil Launches Fern 0.1, Its First Model, With 10M-Token Context
Mirendil released Fern 0.1, its first model, pretrained from scratch with a 10M-token context window. It targets agentic AI R&D work, scores 100% on RULER single needle at 10M, but trails frontier models on agentic coding. The post Mirendil Launches Fern 0.1, Its First Model, With…
Source:
Marktechpost
Discussion (0)