Update README.md

README.md

@@ -11,7 +11,7 @@
   <a href="https://tinker.thinkingmachines.ai/playground">Launch Blog</a>
 </p>
 
-Ricursive Intelligence introduces **Ripple-65B Preview**, a 65B parameter dense Transformer model designed for agentic tasks. 
+Ricursive Intelligence introduces **Ripple-65B Preview**, a 65B parameter dense Transformer model pretrained on 12 trillion tokens with support for up to 1M context length. 
 Ripple-65B Preview delivers strong results on agentic coding, scoring **80.52 on TerminalBench 2.1** and
 surpassing significantly-larger US open-weight models such as 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**). 
@@ -39,7 +39,7 @@ The end result is a model that is especially capable at reasoning and agentic ta
 US open-weight models with more than 10x the number of total parameters.
 At only 65B total parameters, the model can fit on consumer hardware such as the NVIDIA DGX Spark. 
 To encourage local model development, we provide an NVFP4 checkpoint and colab to help you get started. 
-The model is also fast at token generation when running on datacenter GPUs, easily reaching 500+ tok/s on 4xB200 without quantization. 
+The model is also fast at token generation when running on datacenter GPUs, easily reaching 500+ tok/s on 4xB200 prior to quantization or custom kernels. 
 
 **Pre-training.** Ripple-65B Preview is pretrained from scratch on our 12T text-only predominantly-English and code corpus at 8K sequence length. 
 Ripple-65B Base Preview is a **pretrain-only checkpoint** without any midtraining or synthetic QA data.
@@ -102,20 +102,20 @@ Therefore, our results represent a best-effort attempt at a fair comparison.
   | Terminal-Bench 2.1 (Terminus 2) | **80.5** | *51.3* | *55.4* | *79.8* | *67.8* |
   | SWE-bench Verified | — | 71.9 | 77.6 | *78.7* | **81.6** |
   | SWE-bench Pro v1.0 | — | *46.9* | 54.3 | — | **66.1** |
-  | SWE-bench Pro v2.0 Hard | — | *30.7* | *2.0* | *52.3* | 76.5 |
+  | SWE-bench Pro v2.0 Hard | — | *30.7* | *2.0* | *52.3* | **76.5** |
   | DeepSWE 1.1 | — | *0.22* | *2.0* | **60.0** | 43.3 |
   | SciCode | 39.4 | 44.6 | *39.2* | **58.5** | 49.2 | 
-  | τ³-bench banking | —| 22.6 | 23.7 | *14.4* | **36.1** |
+  | τ³-bench banking | 31.27 | 22.6 | 23.7 | *14.4* | **36.1** |
   | AutomationBench (Public) | 36.4 | *8.4* | *14.6* | *42.4* | 39.3 | 
   | MCP Atlas | — | *62.2* | 76.0 | **89.2** | 80.8 |
   | **Safety** | | | | | |
-  | FORTRESS Adversarial | 83.8 | *65.8* | 78.0 | *89.1* | *81.3* |
-  | FORTRESS Benign | 95.5 | *89.1* | 95.9 | *88.1* | *86.4* |
-  | StrongREJECT | 97.5 | *100.0* | 98.6 | *96.6* | *100.0* |
+  | FORTRESS Adversarial | 83.8 | *65.8* | 78.0 | ***89.1*** | *81.3* |
+  | FORTRESS Benign | 95.5 | *89.1* | **95.9** | *88.1* | *86.4* |
+  | StrongREJECT | 97.5 | ***100.0*** | 98.6 | *96.6* | ***100.0*** |
   | **Instruction Following**  | | | | | |
   | IFBench | 79.2 | **81.7** | 79.8 | *78.6* | 79.6 |
   | **Long Context**  | | | | | |
-  | AA-LCR | 75.0 | 65.4 | *78.3* | *80.7* | **78.2** |
+  | AA-LCR | 75.0 | 65.4 | *78.3* | ***80.7*** | 78.2 |
 
 </div>
 
@@ -129,7 +129,7 @@ With the rise of the importance of posttraining, many modern LLM recipes now ble
 For our first model, we kept our pretraining data clean with minimal synthetic data, focusing on high quality natural data rather than aggressively targeting evals.
 We instead added significantly more synthetic data during midtraining, which allowed us to quickly iterate on our synthetic data recipe without blocking pretraining.
 
-Ripple-65B Base Preview is a pretrain-only checkpoint. The model overall compares favorably with Poolside Laguna M.1,
+Ripple-65B Base Preview is a pretrain-only checkpoint. The base model overall beats Poolside Laguna M.1,
 and is comparable with DeepSeek V3.1 and V4 Flash on reasoning, math, and code.
 Nemotron 3 Ultra dominates on traditional pretraining evals, likely due to their heavy use of [task-seeded synthetic data](https://huggingface.co/blog/nvidia/task-seeded-sdg).
 However, its pretraining eval performance did not translate to modern agentic tasks, likely due to their synthetic data being out of distribution relative to these tasks.