# Ripple-65B Preview ![Organization](https://img.shields.io/badge/Org-Ricursive%20Intelligence-555) ![Context](https://img.shields.io/badge/Context-1%2C048%2C576%20tokens-1f6feb) ![Pricing](https://img.shields.io/badge/Price-Open%20weights-2da44e) ![License](https://img.shields.io/badge/License-Apache%202.0-d29922) ![Modality](https://img.shields.io/badge/Modality-Text-8957e5) ![Released](https://img.shields.io/badge/Released-Nov%202026-555) > **Quick answer:** Ripple-65B Preview is Ricursive Intelligence's first public model: a 65B-parameter dense Transformer pretrained on 12 trillion tokens with a 1,048,576-token (1M) context window, released November 6, 2026. It scores 80.52% on TerminalBench 2.1 (self-reported), which the lab reports as US open-weight state-of-the-art on agentic coding — ahead of significantly larger open-weight rivals including Reflection Beam 501B (80.1%), Meta Muse Spark Open 816B (76.4%), Thinking Machines Inkling 975B (63.8%), and NVIDIA Nemotron 3 Ultra 550B (56.4%). It is open-weight under Apache 2.0. ## At a Glance **Where Ripple-65B Preview leads** - 80.52% TerminalBench 2.1 (self-reported) — ahead of every compared open-weight model regardless of size, per the lab's own comparison table - Dense 65B architecture fits on a single NVIDIA DGX Spark, unlike the 500B+ MoE rivals it outperforms on agentic coding - 97.1–98.3% across AIME 2024/2025/2026 — competitive math scores at a fraction of rivals' parameter count - Reported $32M total R&D cost — about one-tenth the compute spend the lab attributes to the prior open-weight SOTA **Where it lags** - 89.6% MMLU and 84.2% MMLU-Pro trail all four larger MoE comparison models (86.8–87.9% MMLU-Pro) - 85.2% GPQA Diamond is the lowest of the five compared models (others: 87.0–90.6%) - FORTRESS Benign (over-refusal) at 95.5% is mid-pack; FORTRESS Adversarial (83.8%) trails Muse Spark Open (94.2%) and Reflection Beam (92.1%) **Best for:** teams wanting strong agentic-coding and terminal-automation performance from a model small enough to self-host on a single workstation-class GPU, rather than a 500B+ parameter MoE. ## What Ripple-65B Preview Is Ripple-65B Preview is the first public model release from Ricursive Intelligence, a lab building AI-driven chip design tools. It is a dense (non-MoE) Transformer pretrained from scratch on 12 trillion tokens, using a "thin-and-deep" aspect ratio (narrow d_model of 5,120 across 80 layers) derived from an Olmo 3 baseline with architectural changes — gated attention, global NoPE, FFN activation clipping, and value residuals — aimed at improving training stability and reasoning at this depth. The lab explicitly chose a dense architecture over MoE to reduce training/serving infrastructure complexity, reporting 40% MFU throughout pretraining and posttraining without custom kernels or quantization. Post-training focused narrowly on agentic coding (midtraining, long-context extension to 2M tokens internally, SFT, and RL), which the benchmark table reflects: Ripple-65B's agentic/coding scores are strongly competitive with or ahead of MoE rivals carrying 8–15x more total parameters, while its general-knowledge scores (MMLU, GPQA) trail those same larger models. Ripple-65B Preview ships alongside a base-model checkpoint (Ripple-65B Base Preview, not separately tracked on this page) and is available in BF16 and NVFP4 numerics, plus third-party hosting via Together AI, Fireworks, GMI Cloud, and OpenRouter. ## Specifications | Field | Value | |---|---| | Organization | Ricursive Intelligence | | Parameters | 65B (dense) | | Context window | 1,048,576 tokens (1M) | | Architecture | Dense Transformer, thin-and-deep (d_model 5,120, 80 layers), gated attention + global NoPE + value residuals | | License | Apache 2.0 | | Release date | November 2026 | | Modality | Text | ## Pricing Open weights, self-hostable — no per-token API pricing from Ricursive Intelligence directly. Available on [Hugging Face](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) in BF16 and NVFP4, and hosted by third parties including [Together AI](https://www.together.ai/models/ripple-65b-preview), [Fireworks](https://fireworks.ai/models/fireworks/ripple-65b-preview), [GMI Cloud](https://console.gmicloud.ai/model/llm/ripple-65b-preview/abfc023e-35b7-4e9e-bb97-cef5db95f171), and [OpenRouter](https://openrouter.ai/ricursiveai/ripple-65b-preview) (check those providers for current per-token rates). ## Context Window Ripple-65B Preview has a **1,048,576-token (1M) context window** — roughly 1,500+ pages of text in a single request. The lab reports an internal long-context training stage extending to 2M-token sequences during posttraining, though the publicly stated max context is 1M. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [MMLU](/benchmarks/hendrycks-et-al/mmlu) | 89.6 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [MMLU-Pro](/benchmarks/wang-et-al/mmlu-pro) | 84.2 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [GPQA Diamond](/benchmarks/rein-et-al/gpqa-diamond) | 85.2 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [AIME 2024](/benchmarks/mathematical-association-of-america/aime-2024) | 97.1 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [AIME 2025](/benchmarks/mathematical-association-of-america/aime-2025) | 98.3 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [AIME 2026](/benchmarks/mathematical-association-of-america/aime-2026) | 95.0 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [HMMT 2026](/benchmarks/harvard-mit-mathematics-tournament/hmmt-2026) | 90.5 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) (Feb 2026 set) | 2026-11 | | [TerminalBench](/benchmarks/tbench/terminalbench-21) | 80.52 | [Ricursive launch blog](https://www.ricursive.com/ripple) (self-reported) | 2026-11 | | [SWE-bench Verified](/benchmarks/swe-bench-verified) | 76.5 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [SWE-bench Pro](/benchmarks/swe-bench-team/swe-bench-pro) | 61.42 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) (v1.0) | 2026-11 | | [DeepSWE](/benchmarks/datacurve/deepswe) | 44.3 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) (v1.1, Terminus 2 harness) | 2026-11 | | [SciCode](/benchmarks/tian-et-al/scicode) | 39.4 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) (Inspect AI harness) | 2026-11 | | [τ³-bench banking](/benchmarks/tau3-banking) | 31.27 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [AutomationBench](/benchmarks/automationbench) | 36.4 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) (public subset) | 2026-11 | | [MCP Atlas](/benchmarks/mcp-atlas) | 71.67 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [FORTRESS](/benchmarks/scale-ai/fortress) (adversarial) | 83.8 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [FORTRESS](/benchmarks/scale-ai/fortress) (benign) | 95.5 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [StrongREJECT](/benchmarks/souly-et-al-uc-berkeley/strongreject) | 97.5 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [IFBench](/benchmarks/ifbench) | 79.2 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | | [AA-LCR](/benchmarks/artificial-analysis/aa-lcr) | 75.0 | [Ripple-65B Preview model card](https://huggingface.co/ricursiveAI/Ripple-65B-Preview) | 2026-11 | *Scores above are reported by Ricursive Intelligence and shown for context. They are not Benchgen measurements. The lab notes Terminal-Bench 2.1 scores shift notably by harness/sequence length — it also reports an 80.5 Terminus-2-harness score (not tabulated here as a separate row; see model card) alongside the 80.52 self-reported figure.* **Not mapped to an existing Benchgen benchmark (bucket C, documented per skill policy rather than guessed):** - **SWE-bench Pro v2.0 Hard** — a harder, dated subset that doesn't match our existing `swe-bench-pro` entity's general methodology; excluded rather than silently widened. - **Base-model pretraining evals** (MMLU, BBH, HellaSwag, etc. for Ripple-65B *Base* Preview) — these describe a distinct pretrain-only checkpoint, not the instruct model this page covers; deferred to a possible future base-model page rather than conflated here. ## Ripple-65B Preview vs Alternatives | Model | Context | TerminalBench 2.1 | GPQA Diamond | Price (in/out per 1M) | |---|---|---|---|---| | **Ripple-65B Preview** | 1M | 80.52 | 85.2 | Open weights | | [Beam 1.0 501B-A23B](/models/reflection/beam-1-0-501b-a23b) | — | 80.1 | 90.6 | Open weights | | [Muse Spark Open 816B-A42B](/models/meta/muse-spark-open-816b-a42b) | — | 76.4 | 89.0 | Open weights | | [Nemotron 3 Ultra 550B-A55B](/models/nvidia/nemotron-3-ultra-550b-a55b) | — | 56.4 | 87.0 | Open weights | Ripple-65B Preview trades roughly a 1–5 point deficit on pure knowledge recall (GPQA, MMLU) for leading agentic-coding throughput at 8–15x fewer total parameters than its closest rivals — a meaningfully cheaper self-hosting footprint for teams whose workload is terminal/coding agents rather than broad knowledge Q&A. ## How Ripple-65B Preview Performs on Real Agent Tasks Ripple-65B Preview's benchmark profile is narrowly optimized: strong on TerminalBench, SWE-bench Verified, MCP-Atlas, and τ³-bench banking — all agentic, tool-use, or terminal-automation evaluations — while trailing on knowledge-heavy tasks. This matches the lab's stated posttraining focus (a relatively light posttraining pass concentrated on agentic coding rather than broad capability). A leaderboard score establishes peak capability under the vendor's own harness and effort settings; it does not establish reliability across a team's specific tools, error-recovery behavior, or cost-at-scale, all of which matter more for production agent deployments than a single aggregate benchmark number. ## Use Ripple-65B Preview via API ```python from openai import OpenAI # swap SDK/base_url per provider, e.g. Together AI, Fireworks, OpenRouter client = OpenAI(api_key="YOUR_API_KEY", base_url="https://api.together.xyz/v1") response = client.chat.completions.create( model="ricursiveai/ripple-65b-preview", messages=[{"role": "user", "content": "Write a shell script to rotate log files older than 7 days."}], ) print(response.choices[0].message.content) ``` ## Related Models & Benchmarks - Compare: [Beam 1.0 501B-A23B](/models/reflection/beam-1-0-501b-a23b) · [Nemotron 3 Ultra 550B-A55B](/models/nvidia/nemotron-3-ultra-550b-a55b) - Benchmarks: [TerminalBench](/benchmarks/tbench/terminalbench-21) · [SWE-bench Verified](/benchmarks/swe-bench-verified) · [GPQA Diamond](/benchmarks/rein-et-al/gpqa-diamond) ## Frequently Asked Questions

What is Ripple-65B Preview?

Ripple-65B Preview is Ricursive Intelligence's first public model release: a 65B-parameter dense Transformer optimized for agentic coding and terminal automation, released November 6, 2026.

What is Ripple-65B Preview's context window?

1,048,576 tokens (1M) — roughly 1,500+ pages of text in a single request.

How much does Ripple-65B Preview cost?

It is open weights — free to download and self-host. Third-party providers (Together AI, Fireworks, GMI Cloud, OpenRouter) host it for per-token API access at their own published rates.

Is Ripple-65B Preview open source?

Yes — Apache 2.0 license, with weights available on Hugging Face in BF16 and NVFP4 numerics.

What is Ripple-65B Preview's knowledge cutoff?

Not disclosed by Ricursive Intelligence as of this writing.

--- *Specs and scores sourced from Ricursive Intelligence's official Hugging Face model card and launch blog; third-party benchmark scores attributed inline. Last updated 2026-11-07.*

Ripple-65B Preview — Evaluations

Evaluations