Size
65B
65B active — only part of the model works on each answer, but all of it has to fit in memory
Ripple-65B Preview
Context
1M
the longest conversation the model is built to hold
Lab
Ricursive Intelligence
the lab that released it
Released
Nov 5, 2026
Licence: Apache 2.0 — free use, modification, commercial use and redistribution; no separate agreement needed.
Worth knowing
Ricursive Intelligence's first model, released Nov 4, 2026: a dense 65B-parameter Transformer pretrained from scratch on 12T tokens, with context up to 1M tokens and a self-reported 80.52 on Terminal-Bench 2.1 (Ricursive's figures, not independently reproduced). Checkpoints: BF16 instruct (about 130GB), NVFP4 and BF16 base, all Apache 2.0 and ungated on Hugging Face. At 65B the BF16 weights alone are about 130GB — two 80GB cards or one 141GB card; the NVFP4 build is the single-card route (Ricursive says it fits DGX Spark-class hardware).
What it needs
Full precision (FP16)
130 GB
the weights, uncompressed
8-bit
65 GB
the weights, compressed to 8-bit
4-bit
33 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.
Runs Ripple-65B Preview at full precision — its 130GB of weights fit this card's 192GB, with room left for the conversation memory.
Frontier training money. If your job fits on anything cheaper, a B200 is budget burned for a spec sheet.
Runs Ripple-65B Preview at full precision — its 130GB of weights fit this card's 288GB, with room left for the conversation memory.
Very costly for small jobs — best value only on big work.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 95GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
This is the Max-Q version: it runs at lower power and is slower than the full card.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 80GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
An older generation — no FP8, so the newer speed tricks for serving don't apply.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 96GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
No NVLink scale-out; strong all-round, but the Max-Q version (listed separately) runs slower.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 80GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
No NVLink, and lower power than the SXM — slower when cards must work together.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 80GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Pays off only when the card stays busy; on bursty work you pay H100 rent for idle silicon.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 94GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Sold as linked pairs, not as a lone card.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 80GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
One provider's label, published alongside its plainer “H100 PCIe” — check which card you are actually booking.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 141GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
More card than most jobs need; the main trap is paying for it on small work.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 80GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Pays off only when kept busy — avoid it for bursty development work.
Runs Ripple-65B Preview compressed — the 8-bit build needs 65GB, which fits this card's 96GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Its processor is ARM, not the usual x86 — some software needs changes to run on it.
Runs Ripple-65B Preview compressed — the 4-bit build needs 33GB, which fits this card's 48GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
An older generation — no FP8, so the newer speed tricks for serving don't apply.
Runs Ripple-65B Preview compressed — the 4-bit build needs 33GB, which fits this card's 48GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
70B at FP16 does not fit, and no flag fixes that. Weak at double-precision maths; no NVLink.
Runs Ripple-65B Preview compressed — the 4-bit build needs 33GB, which fits this card's 48GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
Weak at double-precision maths — not the card for scientific computing. No NVLink.
Runs Ripple-65B Preview compressed — the 4-bit build needs 33GB, which fits this card's 40GB. The compromise: slightly lower answer quality and a shorter conversation memory than the full-precision build.
The 40GB fills up quickly on larger models.
The cheapest verified way to run it
RTX A6000 at Thunder Compute — $0.35/hr — runs it compressed — the 4-bit build.
Launch Thunder Compute ↗ See the RTX A6000 →
Checked Nov 15, 2026
Select a GPU See these GPUs on the prices board, ranked for this model.