Skip to content
Trending storyDeveloping

Mirendil releases Fern 0.1, a from-scratch model with up to 10M-token context

2 articles2 sourcessince Oct 26Last article

Overview

AISummary of 2 articles

Mirendil releases Fern 0.1, its first model, pretrained from scratch with up to 10M tokens of context, available through hosted early access only.

The company says the window equals about 23 MB of code, roughly 500,000 lines, and that Fern 0.1 received only a short round of post-training. It also says next-token prediction keeps improving as context grows to 10M tokens, but the announcement includes no benchmark results for that claim.

Mirendil's own table shows Fern 0.1 scoring 100% on RULER single-needle retrieval at every length from 32K to 10M tokens. On Terminal Bench 4.0 it scores 6.1%, against 66.4% for Claude Opus 5.5, according to MarkTechPost, which reports the figures from Mirendil's table. No open weights or quantized builds have been announced.

Written by AI from the articles below · updated Oct 27, 9:36 AM ET

Check the sources:

Developments

2 developments

  1. Oct 27, 8:57 AM ET · 1 article
    Mirendil releases Fern 0.1, its first coding model with 10M-token context
    MarkTechPost: Mirendil releases Fern 0.1, its first model with a 10M-token context window
  2. Oct 26, 11:05 PM ET · 1 article
    Mirendil releases Fern, a model pretrained from scratch with up to 10M-token context
    meng shao: Mirendil releases Fern, a model pretrained from scratch with 10M-token context

Article timeline

The articles in this story. Times are ET.

Oct 27
  1. MarkTechPostNews
    Mirendil releases Fern 0.1, its first model with a 10M-token context window

    AIMirendil has released Fern 0.1, its first AI model, a from-scratch coding model with a context window of up to 10M tokens, available only through early access. On RULER single-needle retrieval it scores 100% at every length from 32K to 10M tokens, but it scores 6.1% on Terminal Bench 4.0 against 66.4% for Claude Opus 5.5 in Mirendil's own table.

Oct 26
  1. meng shaoX
    Mirendil releases Fern, a model pretrained from scratch with 10M-token context

    AIMirendil releases Fern 0.1, its first model, pretrained from scratch with up to 10M tokens of context. The company says that is about 23 MB of code, roughly 500,000 lines, or several days of agent work held at once. Fern 0.1 received only a short round of post-training, and Mirendil says next-token prediction keeps improving as context grows to 10M.

    Image from @shao__meng's post

Heat Trend

Current comparable heat
10
Peak comparable heat
18Oct 27, 9:00 AM ET
Change in past 24 hours
+16%

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.