Ricursive drops Ripple-65B Preview
Ricursive Intelligence released Ripple-65B Preview, a 65B dense open-weights model trained on 12 trillion tokens with claimed support for up to 1 million-token contexts. The company reports an 80.52 score on Terminal-Bench 2.1, positioning Ripple as a strong contender for agentic coding.
Ripple is one of the more compelling open-weight releases this year, but its benchmark lead remains a company-reported claim awaiting independent validation.
- –Apache 2.0 weights make the model commercially usable and self-hostable.
- –The 1M-token context is an architectural capability; initial hosted access appears limited to 16K tokens.
- –Its dense design could offer more predictable inference than massive mixture-of-experts models.
- –A 65B checkpoint still requires substantial hardware, though quantized versions improve local deployment prospects.
- –Terminal-Bench 2.1 results are promising, but cross-benchmark comparisons should be treated cautiously.
DISCOVERED
1h ago
2026-11-04
PUBLISHED
2h ago
2026-11-04
RELEVANCE
AUTHOR
mark_k