On Oct 26, Mirendil introduced Fern, a from-scratch model with up to 10M-token context.
On Oct 26, Mirendil introduced Fern, a from-scratch model with up to 10M-token context.
Mirendil published this on Oct 26, 2026 (source: mirendil.com).
Mirendil announced Fern, its first model pretrained from scratch with a context window of up to 10 million tokens. Fern 0.1 is the smallest in a planned 0.1–0.5T-parameter family after a short post-training pass; larger siblings are still pretraining. Access is early-access only.
Mirendil published this on Oct 21, 2026 (source: mirendil.com).
Mirendil says agents wrote over 99% of the code for Mirendil-Base while one researcher directed the work, with all-in compute under $10 million. The model is in the 100B-500B parameter range and trails Kimi-K3 on Mirendil's own code benchmark; larger versions are still training and post-training has started.
Items older than 72 hours never go in Top stories and show their original date.