Key points
- Ricursive Intelligence released Ripple-65B Preview with openly available weights and 65 billion parameters.
- The company says the model led open US models on the TerminalBench 2.1 coding benchmark, with a reported total cost of $32 million.
- On other benchmarks, Ripple-65B Preview performed considerably worse than the Western model Inkling 2 Small.
Context
For you, the release adds an open-weight option for evaluating agentic coding systems, but its benchmark results are mixed. The leading TerminalBench 2.1 score and $32 million reported cost are offset by weaker performance elsewhere, making direct comparisons important for technical decision-makers.
What this means for you
Directly applicable
You can use the benchmark claim and reported cost when comparing open-weight coding models, while checking the model's weaker results on other benchmarks.
Who benefits: Developers evaluating open-weight models for agentic programming gain another candidate for practical comparisons and coding experiments.
Share article: