MODELS →

Sage S2.1 Flash

Size
321.3B
18B active — only part of the model works on each answer, but all of it has to fit in memory
Context
1M
the longest conversation the model is built to hold
Lab
InternLM
the lab that released it
Released
Nov 15, 2026

Licence: MIT — free for any use, including commercial. Just keep the licence notice.

Worth knowing

Shanghai AI Laboratory's scientific multimodal agent model, weights on Hugging Face Nov 15, 2026 under MIT: 321.3B total parameters by the repository's own file metadata (the maker describes it as 320B), 18B active per token, 45 layers, 288 routed experts, 1,048,576-token maximum context in its own config. BF16 payload 642.8 GB across 129 files; an FP8 build is published at about 321 GB. Its Sage SciPerception layer reads 10+ scientific input types (molecular graphs, crystal structures, protein sequences and 3D structures, spectra, XRD, ECG, EEG, light curves). The maker's deployment guides (vLLM / SGLang / LMDeploy) use an 8 × H200 tensor-parallel reference setup and require trust_remote_code. All benchmarks on the card are the maker's own — MLE-bench Lite 90.3 is its headline — with no independent replication yet.

What it needs

Full precision (FP16)
643 GB
the weights, uncompressed
8-bit
321 GB
the weights, compressed to 8-bit
4-bit
161 GB
the weights, compressed to 4-bit

How we work this out: memory = the model's total parameters × bytes per weight — 2 bytes at full precision (FP16), 1 byte at 8-bit, half a byte at 4-bit. Weights only — add room for the conversation memory (KV cache) and the serving software. A GPU counts as running a build when its memory covers the weights plus about 20% headroom. Real figures move with context length and software.

GPUs that can run it

Ranked by how well the model runs, then by price.

1.B300 Most recommended Runs it compressed288GB

Runs Sage S2.1 Flash compressed — the 4-bit build needs 161GB, which fits this card's 288GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.

Very costly for small jobs — best value only on big work.

Lowest verified: Hyperstack — $7.40/hr Checked Nov 21, 2026 Launch ↗

At full precision, Sage S2.1 Flash needs more than one GPU — no single card holds its 643GB of weights plus room to work. Every card above runs a compressed build.

The cheapest verified way to run it

B300 at Hyperstack — $7.40/hr — runs it compressed — the 4-bit build.

Launch Hyperstack ↗  See the B300 →

Checked Nov 21, 2026

Select a GPU  See these GPUs on the prices board, ranked for this model.