{
  "@context": "https://schema.org",
  "@id": "https://aggr.aymericbeaumet.com/items/hnrss-org/2026-11-04-ripple-65b-preview-u-s-open-weight-sota-on-agentic-coding/#webpage",
  "@type": [
    "WebPage",
    "ArchiveComponent"
  ],
  "articleBody": "Rather than investing in a brute-force RL climb, we carefully iterated on our SFT and RL recipes with a focus on compute efficiency. We opted not to buy any posttraining data, instead devising our own synthetic data pipelines, quality filtering, and coverage metrics.\n\nOur mixed RL recipe uses [truncated importance sampling](https://openreview.net/forum?id=xsonlP8DRV), [score centering](https://arxiv.org/abs/2609.20807), and [linear length penalty](https://microsoft.ai/pdf/mai-thinking-1.pdf) with high-quality long-horizon synthetic environments. The RL tasks are tuned for our model capabilities, and we continuously added harder tasks as the model improved.\n\nThe end result was a steep RL climb from a strong initial SFT checkpoint. We reached a TerminalBench 2.1 score of 80.5 with only 1.4M rollouts, roughly 70x fewer rollouts than Reflection Beam, the prior SOTA.\n",
  "authors": [
    "Philpax"
  ],
  "captured_at": "2026-11-04T21:49:17.908350494+00:00",
  "category": "news",
  "content_html": "<p>Rather than investing in a brute-force RL climb, we carefully iterated on our SFT and RL recipes with a focus on compute efficiency. We opted not to buy any posttraining data, instead devising our own synthetic data pipelines, quality filtering, and coverage metrics.</p>\n<p>Our mixed RL recipe uses <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://openreview.net/forum?id=xsonlP8DRV\">truncated importance sampling</a>, <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://arxiv.org/abs/2609.20807\">score centering</a>, and <a target=\"_blank\" rel=\"noopener noreferrer\" href=\"https://microsoft.ai/pdf/mai-thinking-1.pdf\">linear length penalty</a> with high-quality long-horizon synthetic environments. The RL tasks are tuned for our model capabilities, and we continuously added harder tasks as the model improved.</p>\n<p>The end result was a steep RL climb from a strong initial SFT checkpoint. We reached a TerminalBench 2.1 score of 80.5 with only 1.4M rollouts, roughly 70x fewer rollouts than Reflection Beam, the prior SOTA.</p>\n",
  "content_markdown": "Rather than investing in a brute-force RL climb, we carefully iterated on our SFT and RL recipes with a focus on compute efficiency. We opted not to buy any posttraining data, instead devising our own synthetic data pipelines, quality filtering, and coverage metrics.\n\nOur mixed RL recipe uses [truncated importance sampling](https://openreview.net/forum?id=xsonlP8DRV), [score centering](https://arxiv.org/abs/2609.20807), and [linear length penalty](https://microsoft.ai/pdf/mai-thinking-1.pdf) with high-quality long-horizon synthetic environments. The RL tasks are tuned for our model capabilities, and we continuously added harder tasks as the model improved.\n\nThe end result was a steep RL climb from a strong initial SFT checkpoint. We reached a TerminalBench 2.1 score of 80.5 with only 1.4M rollouts, roughly 70x fewer rollouts than Reflection Beam, the prior SOTA.\n",
  "dateCreated": "2026-11-04T21:49:17.908350494+00:00",
  "date_created": "2026-11-04T18:53:29+00:00",
  "date_modified": "2026-11-04T18:53:29+00:00",
  "date_published": "2026-11-04T18:53:29+00:00",
  "description": "Rather than investing in a brute-force RL climb, we carefully iterated on our SFT and RL recipes with a focus on compute efficiency. We opted not to buy any posttraining data, instead devising our own synthetic data pipelines, quality…",
  "external_url": "https://www.ricursive.com/ripple",
  "git": {
    "commit": "20068632b65d2830a2d54f25ca1d90d697bea2c0",
    "path": "items/hnrss-org/2026/11/2026-11-04-ripple-65b-preview-u-s-open-weight-sota-on-agentic-coding.md",
    "permalink": "https://github.com/aymericbeaumet/aggr-instance/blob/20068632b65d2830a2d54f25ca1d90d697bea2c0/items/hnrss-org/2026/11/2026-11-04-ripple-65b-preview-u-s-open-weight-sota-on-agentic-coding.md"
  },
  "id": "urn:aggr:item:391569679cd3557c37c2d8e83e53f0dcc15dda4d",
  "isBasedOn": {
    "@id": "https://www.ricursive.com/ripple"
  },
  "mainEntity": {
    "@id": "https://www.ricursive.com/ripple",
    "@type": "CreativeWork",
    "archivedAt": {
      "@id": "https://aggr.aymericbeaumet.com/items/hnrss-org/2026-11-04-ripple-65b-preview-u-s-open-weight-sota-on-agentic-coding/#webpage"
    },
    "author": [
      {
        "@type": "Person",
        "name": "Philpax"
      }
    ],
    "dateModified": "2026-11-04T18:53:29+00:00",
    "datePublished": "2026-11-04T18:53:29+00:00",
    "headline": "Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding",
    "image": {
      "@type": "ImageObject",
      "caption": null,
      "height": 90,
      "url": "https://framerusercontent.com/images/u4OXzV0utwAtaPgcVxQtTXmkg.png",
      "width": 160
    },
    "timeRequired": "PT1M",
    "url": "https://www.ricursive.com/ripple",
    "wordCount": 120
  },
  "name": "Archived snapshot: Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding",
  "preview": {
    "alt": null,
    "color": null,
    "height": 90,
    "placeholder": {
      "data_url": "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII=",
      "hash": ""
    },
    "url": "https://framerusercontent.com/images/u4OXzV0utwAtaPgcVxQtTXmkg.png",
    "width": 160
  },
  "reading_minutes": 1,
  "replicated_at": null,
  "source": "hnrss-org",
  "source_name": "Hacker News: Front Page",
  "summary": null,
  "tags": [],
  "temporalCoverage": "2026-11-04T21:49:17.908350494+00:00",
  "title": "Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding",
  "url": "https://aggr.aymericbeaumet.com/items/hnrss-org/2026-11-04-ripple-65b-preview-u-s-open-weight-sota-on-agentic-coding/",
  "word_count": 120
}