ripple-preview:65b-mlx-bf16

248 Downloads Updated 4 hours ago

65B dense model built for agentic coding, with up to 1M tokens of context.

tools thinking 65b

ollama run ripple-preview:65b-mlx-bf16

curl http://localhost:11434/api/chat \
  -d '{
    "model": "ripple-preview:65b-mlx-bf16",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='ripple-preview:65b-mlx-bf16',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'ripple-preview:65b-mlx-bf16',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 4 hours ago

4 hours ago

0510c48c0544 · 138GB

json

{ "architectures": [ "RicursiveForCausalLM" ], "auto_map": { "AutoConfig": "configuration_ricursive.

3.5kB

json

{"bos_token": {"content": "<|endoftext|>", "lstrip": false, "normalized": false, "rstrip": false, "s

382B

json

{ "version": "1.0", "truncation": null, "padding": null, "added_tokens": [ { "id": 100256, "content"

7.1MB

json

{ "add_prefix_space": false, "added_tokens_decoder": { "100256": { "content": "<|extra_id_0|>", "lst

4.6kB

json

{ "architectures": [ "DFlash2DraftModel" ], "attention_bias": false, "attention_dropout": 0.0, "bos_

1.3kB

params

{ "num_ctx": 16384 }

18B

license

Apache-2.0

10B

1284 tensors

138GB

Readme

Ripple-65B Preview is a 65B dense model from Ricursive Intelligence, built for agentic coding. It is a general-purpose version of the model behind Ricursive’s AI chip-design platform.

Get started

ollama run ripple-preview

Key Features

Agentic coding: trained with large-scale reinforcement learning to work like a coding agent, carrying out multi-step tasks in a terminal.
Dense, trained from scratch: 65B parameters, all active on every token, pretrained by Ricursive on 12 trillion tokens.
Long context: up to 1M tokens, enough for large codebases and long agent sessions.
Faster generation: on Ollama’s MLX engine, Ripple uses DFlash 2 speculative decoding to generate faster..
Open license: Apache 2.0, commercial use allowed.

Benchmarks

Ricursive reports a score of 80.5 on Terminal-Bench 2.1. See the launch post for details.

Reference

Ricursive launch post · Hugging Face card

![images.png](/assets/library/ripple-preview/c905c99a-002b-40e7-bb5e-2d8666a77011)

Ripple-65B Preview is a 65B dense model from Ricursive Intelligence, built for agentic coding. It is a general-purpose version of the model behind Ricursive's AI chip-design platform.

### Get started

```
ollama run ripple-preview
```
### Key Features
- **Agentic coding**: trained with large-scale reinforcement learning to work like a coding agent, carrying out multi-step tasks in a terminal.
- **Dense, trained from scratch**: 65B parameters, all active on every token, pretrained by Ricursive on 12 trillion tokens.
- **Long context**: up to 1M tokens, enough for large codebases and long agent sessions.
- **Faster generation**: on Ollama’s MLX engine, Ripple uses DFlash 2 speculative decoding to generate faster..
- **Open license**: Apache 2.0, commercial use allowed.

### Benchmarks
Ricursive reports a score of 80.5 on Terminal-Bench 2.1. See the [launch post](https://www.ricursive.com/ripple) for details.

### Reference

[Ricursive launch post](https://www.ricursive.com/ripple) · [Hugging Face card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview)

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)