Skip to content

NewsRAM

Ripple-65B Preview Arrives in Ollama: What PC Do You Need to Run It?

Ricursive Intelligence's Ripple-65B Preview, a dense 65-billion-parameter Apache 2.0 coding model with a reported 80.52 TerminalBench 2.1 score and up to 1 million tokens of context, is now in Ollama's library as ripple-preview, a 39-46GB download. A typical 8GB or 16GB gaming GPU will not hold it in VRAM on its own.

By , Hardware News & Market Editor

4 min read

Ripple-65B Preview Arrives in Ollama: What PC Do You Need to Run It?

A new 65-billion-parameter model has landed in Ollama's library, and it is aimed at people who want a coding agent on their own machine rather than in the cloud. Ripple-65B Preview comes from Ricursive Intelligence, and the company's pitch is unusually specific: it says the model was trained from scratch for agentic coding, scores 80.52 on TerminalBench 2.1, and is released under an Apache 2.0 licence. The practical question for a UK PC owner is simpler. Can your machine actually run a 39-46GB download?

The figures and commands below come from Ricursive's official announcement and from Ollama's library page, both read on 14 November 2026.

What Ripple-65B Preview is

Ricursive describes Ripple-65B Preview as a general-purpose, open-weight variant of the model that powers its own AI chip-design platform. Official details it gives are:

  • A dense 65-billion-parameter transformer (every parameter is used on every token, not a mixture-of-experts design)

  • Pretrained on 12 trillion tokens, with a context window of up to 1 million tokens

  • Released under Apache 2.0, so commercial use is allowed on those terms

  • A reported 80.52 score on TerminalBench 2.1, which Ricursive calls US open-weight state of the art for agentic coding

Ollama's model card matches the 65B dense description, the 1M-token context in its key features list, the Apache 2.0 licence, and a rounded 80.5 Terminal-Bench 2.1 figure attributed to Ricursive. It also lists a 39GB to 46GB download for the ripple-preview and ripple-preview:65b tags. Treat that as the amount of disk you need, not as a precise VRAM requirement: how much actually sits in the graphics card depends on the quantisation and on whether parts of the model are offloaded to system RAM.

How to run it

If you already use Ollama, the official command is:

ollama run ripple-preview

The first run downloads the model. You can then chat in the terminal, or point other tools at Ollama's local API, as we describe in How to use Ollama's OpenAI-compatible API. Ollama's library page also shows launch commands for several coding tools. If you are new to Ollama, start with our Ollama or LM Studio guide before you pull a file this large.

The weights are also listed on Hugging Face and on several cloud hosts, according to Ricursive. This article is about running them locally.

What kind of PC you need

A typical 8GB or 16GB gaming GPU will not hold a 39-46GB model in video memory on its own. You would need a lot of system RAM and a willingness to offload, which is slower, or a machine with a very large unified memory pool. Ricursive itself says the model's "small parameter envelope" is meant to run efficiently on local devices such as the NVIDIA DGX Spark, which have a lot of on-package memory. We compared that class of machine with AMD's high-memory AI PCs in DGX Spark vs Ryzen AI Max+ 395, and we covered the desktop units in NVIDIA RTX Spark mini PCs.

For a more typical desktop, our GPU for local models and how much RAM for local AI guides are the realistic shopping lists. A 12GB or 16GB card that is enough for mid-size local models is not automatically enough for a 65B dense model. Check Ollama's own memory behaviour after the download, rather than assuming the file will sit entirely on the GPU.

Disk space matters too. You need tens of gigabytes free, preferably on a fast NVMe drive; our best SSD guide is the place to start if the boot drive is already full.

How to read the benchmark claim

Ricursive's TerminalBench number is the company's own reported result, with comparisons to other US open-weight models that it names on the same page. It is not a Forge PC test, and it is not a substitute for trying the model on the work you actually do. Agentic coding scores also depend on the tools and the exact harness. If the number is what attracted you, treat it as a vendor claim and judge the model on your own repo once it loads.

Our verdict

Ripple-65B Preview is a genuinely large, Apache 2.0 coding model that you can now pull with ollama run ripple-preview. Officially it is a 65B dense model with up to 1M tokens of context, and Ollama's library lists a 39-46GB download. That puts it out of easy reach for a normal 8GB or 16GB gaming GPU; a high-memory workstation, a DGX Spark-class machine, or a heavy RAM offload is the realistic path.

The bottom line

Ripple-65B Preview is a genuinely large, Apache 2.0 coding model that you can now pull with `ollama run ripple-preview`. Officially it is a 65B dense model with up to 1M tokens of context, and Ollama's library lists a 39-46GB download. That puts it out of easy reach for a normal 8GB or 16GB gaming GPU; a high-memory workstation, a DGX Spark-class machine, or a heavy RAM offload is the realistic path.

Sources

  1. 01Ricursive Intelligence, Introducing Ripple-65B Preview, a Frontier Open-Weight Model
  2. 02Ollama, ripple-preview library (read 14 November 2026)

About the author

Eleanor Shaw

Hardware News & Market Editor

Eleanor covers daily PC hardware news, silicon roadmaps, driver releases and UK retail pricing trends across AMD, Intel and Nvidia. She tracks board-partner specifications, industry standards bodies (PCI-SIG, JEDEC, VESA) and supply-chain shifts affecting British PC builders.

More from Eleanor Shaw

Read next

Patriot Launches AI Matrix SSDs and DDR5 Kits Pitched at Local AI PCs

News

Patriot Launches AI Matrix SSDs and DDR5 Kits Pitched at Local AI PCs

Patriot's AI Matrix range pairs MS560 (Gen5), MS450 (Gen4) and MS350 (Gen3) M.2 SSDs with DDR5 kits in PRO, Mid and Entry tiers matched to local LLM sizes. The SSDs carry a two-year warranty, DRAM specs aren't fully published and there's no UK price yet.

Eleanor Shaw3 min

Best 128GB DDR5 RAM Kit in 2026: Five 2x64GB Kits Compared

Best Picks

Best 128GB DDR5 RAM Kit in 2026: Five 2x64GB Kits Compared

Five 2x64GB DDR5 kits compared on their makers' specs: the G.SKILL Flare X5 DDR5-6000 CL34 is our pick for Ryzen, with the Trident Z5 Neo RGB for lighting, the Ripjaws S5 DDR5-6400 for Intel and the Crucial Pro DDR5-5600 if you'd rather not overclock.

Hannah Vance4 min