Frontier Log

Ricursive Intelligence

Ripple-65B Preview

Released · ricursive.com

Context window

TokensContext window
Default1M
Context history
DateContext window
ricursive.com · 2 · Corroborated1M

Benchmark results

BenchmarkScore
SWE-bench VerifiedOpenCode harness; maximum effort76.5%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
SWE-bench ProSet not stated · v1.0; OpenCode harness; maximum effort61.42%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
GPQA Diamondmaximum effort85.2%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
AutomationBench, no release namedPublic set · OpenCode harness and maximum effort per the model card36.4%Corroborated The same number is printed in 1 other lab document: 1
DeepSWE v1.1Terminus 2 harness and maximum effort per the model card44.3%Corroborated The same number is printed in 1 other lab document: 1
AIME 2024No tools · maximum effort; tools not stated97.1%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
AIME 2025No tools · maximum effort; tools not stated98.3%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
Terminal-Bench 2.1harness not stated; maximum effort per the model card80.52%Corroborated The same number is printed in 1 other lab document: 1
MMLU-Promaximum effort84.2%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
MCP AtlasSet not stated · OpenCode harness; maximum effort71.67%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
IFBenchPrompt-level or unstated · maximum effort79.2%Checked A person checked this number against its source on 7 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.
Sources

Reported by Ricursive Intelligence

Related

All models