763B
- input
- $0.20
- cached
- $0.05
- output
- $0.60
- Context
- 1Mtokens
- Released
- 1w ago
- Size
- 65Bparameters
Cost/1M tokens
Ripple-65B Preview is the first model from Ricursive Intelligence, a 65B dense Transformer trained from scratch on 12 trillion tokens with up to 1M tokens of context. It is released with open weights under Apache 2.0 and aimed at agentic coding, where Ricursive reports 80.52 on TerminalBench 2.1.
- Language model
- Open weights · 65B
- 1M context
- Reasoning
- Tools
- Structured output
How it compares
Frequently asked questions
Running locally
By our estimate, Ripple-65B Preview needs about 45.8 GB at Q4, 79.9 GB at Q8 and 150 GB at FP16 of VRAM. That covers the weights plus headroom for a short context, and longer contexts need more.
No, by our estimates Ripple-65B Preview does not fit fully on a 24 GB GPU even at Q4. It needs more VRAM or offloading to system memory.
Yes, a Mac with at least 64 GB of unified memory should run Ripple-65B Preview at Q4, which needs an estimated 45.8 GB. This assumes macOS gives the GPU about 75% of unified memory.
API and pricing
Ripple-65B Preview is listed at $0.20 per 1M input tokens and $0.60 per 1M output tokens. Input prices across 3 providers range from $0.10 to $0.20 per 1M tokens.
PrimeIntellect has the lowest blended price for Ripple-65B Preview (3:1 input to output), at $0.10 per 1M input tokens and $0.30 per 1M output tokens.
Yes, cached input for Ripple-65B Preview costs $0.05 per 1M tokens, compared with $0.20 for uncached input, a 75% discount.
A request with 10,000 input tokens and 1,000 output tokens costs about $0.0013 with Ripple-65B Preview at the cheapest provider, PrimeIntellect.
2 of 3 tracked Ripple-65B Preview endpoints had 99% uptime or better over the last day. The best was Together at 100.0%.
Privacy and data
PrimeIntellect and Together offer zero data retention for Ripple-65B Preview, so prompts are not stored after the request.
Yes, Ripple-65B Preview has open weights, so you can run it on your own hardware and keep prompts on your own machines.
Capabilities and specs
Ripple-65B Preview has a 1M token context window. It can write up to 131K output tokens per response.
No, Ripple-65B Preview takes text input only.
Yes, Ripple-65B Preview supports tool calling (function calling), so it can request calls to tools you define.
Yes, Ripple-65B Preview supports structured output, so responses can be constrained to JSON.
Yes, Ripple-65B Preview has a reasoning mode. 2 of 3 tracked providers expose it.
Quality and benchmarks
Similar models from other labs include Nemotron 3.5 Super VL (NVIDIA), Beam 1.0 501B A23B (Reflection AI) and GLM 5.3 (Z.ai). They are the closest matches by kind, price and release date.
Nemotron 3.5 Super VLRipple-65B Preview vs Nemotron 3.5 Super VLBeam 1.0 501B A23BRipple-65B Preview vs Beam 1.0 501B A23BGLM 5.3Ripple-65B Preview vs GLM 5.3
Licensing and access
Ripple-65B Preview has open weights under the Apache 2.0 license, so you can download and run it yourself.
Yes, Ripple-65B Preview is released under the Apache 2.0 license, which allows commercial use. Check the license text for attribution terms.
The Ripple-65B Preview weights are on Hugging Face at ricursiveAI/Ripple-65B-Preview.
Basics
Ripple-65B Preview was made by Ricursive Intelligence. It was released on November 4, 2026.
Yes, Ripple-65B Preview has open weights, so you can run it on your own hardware without API fees. Paid API access is listed at $0.20 per 1M input tokens and $0.60 per 1M output tokens.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.