AMD open-sources Instella 16B MoE with checkpoints from pretraining to RL
bronzeagepapi · x · 2026-07-25
AMD releases Instella 16B MoE as a fully open foundation model
AMD says Instella 16B MoE is its first MoE foundation model and its first built with FarSkip architecture.
- Checkpoints are published after every stage: pretraining, long-context, SFT, DPO, and RL.
- The team also released dataset details, training recipes, and code for reproducibility.
- The post frames this as a fully open release rather than a partial demo or teaser.
More from Models
- Fern's "context is intelligence" thesis: prediction keeps improving out to 10.2M tokens — bneyshabur · 2026-10-26
- Mirendil launches Fern, a from-scratch pretrained model with up to 10M-token context — bneyshabur · 2026-10-26
- Users spot a dedicated translation UI appearing in Claude — BLCNYY · 2026-10-26
- GrokBot user: flight booking and shopping all blocked, agent can't do anything useful — lucasmeijer · 2026-10-26
- Vending-bench-style eval: Sol 5.6 appears the most aggressive contract-violator — xuanalogue · 2026-10-26
- Fable 5.5 launch rumors persist for third week, but no frontier model drops this month — patience_cave · 2026-10-26