Darwin-180B-RSI
PulseAugur coverage of Darwin-180B-RSI — every cluster mentioning Darwin-180B-RSI across labs, papers, and developer communities, ranked by signal.
- developed Zenkit 90%
- developed by POCKET-Darwin-180B-GGUF 90%
- developed by LEXam-hard 90%
- developed by POCKET-Darwin-180B 90%
- instance of SuperCLUE 90%
- instance of Darwin-180B-RSI-R3 90%
- competes with Kimi-k3.1-pro 80%
- used by POCKET-Darwin-180B-GGUF 70%
- used by LEXam-hard 70%
- affiliated with SuperCLUE 70%
- instance of LEXam-hard 70%
- instance of Bidraft 70%
- 2026-11-09 research_milestone VIDRAFT's Darwin-180B-RSI model achieved top scores on the LEXam and LEXam-hard legal reasoning benchmarks. source
- 2026-11-06 research_milestone VIDRAFT's Darwin-180B-RSI model was updated using a novel method that modified only 0.02% of its parameters and demonstrated efficient inference on consumer hardware. source
- 2026-11-05 research_milestone Darwin-180B-RSI achieved the top rank among open-weight models in math and science reasoning on the SuperCLUE benchmark. source
- 2026-09-30 product_launch VIDRAFT released the Darwin-180B-RSI model, which achieved top rankings on five Hugging Face leaderboards. source
- 2026-09-29 product_launch VIDRAFT released Darwin-180B-RSI, a 180-billion-parameter reasoning model. source
10 day(s) with sentiment data
Darwin-180B-RSI demonstrates significant domain-agnostic reasoning
Darwin-180B-RSI has achieved top performance on diverse benchmarks, including STEM, legal reasoning, and document processing, without domain-specific fine-tuning. This suggests its recursive self-improvement (RSI) training methodology is highly effective at developing generalizable reasoning skills, a notable advancement in LLM training.
VIDRAFT to release smaller, quantized variants of Darwin-180B-RSI
The successful deployment of the POCKET-Darwin-180B-GGUF variant on consumer hardware indicates a strategic focus on accessibility. It is likely that VIDRAFT will continue to release further optimized, smaller, or more heavily quantized versions of Darwin-180B-RSI to cater to a wider range of devices and use cases.
Darwin-180B-RSI to release enterprise-focused API within 60 days
Given Darwin-180B-RSI's strong performance on enterprise-relevant benchmarks like document parsing and field extraction, and its ability to run on consumer hardware, it's plausible VIDRAFT will offer an enterprise API to monetize these capabilities. This would allow businesses to integrate its advanced reasoning and data extraction features without managing local deployments.
-
Darwin Lab offers secure, on-premise AI for sensitive research data
Darwin Lab, developed by VIDRAFT, offers a private, on-premise AI research platform designed to handle sensitive data without internet connectivity. The platform integrates the Darwin-180B-RSI model, which has demonstra…
-
Open-weight AI model Darwin-180B-RSI outperforms larger rivals in reasoning tests
A new open-weight AI model named Darwin-180B-RSI, developed by Korea's VIDRAFT, has achieved a score of 73.68 in mathematical and scientific reasoning on the SuperCLUE evaluation. This performance places it second overa…
-
Darwin-180B-RSI: 180B parameter model tops open-source LLM rankings
Vidraft Lab's open-source large language model, Darwin-180B-RSI, has achieved the top rank among open-weight models and second overall in the SuperCLUE November evaluation for math and science reasoning. This 180-billio…
-
VIDRAFT's Darwin-180B-RSI model leads legal reasoning benchmarks without specific training
VIDRAFT's Darwin-180B-RSI model has achieved top rankings on the LEXam and LEXam-hard legal reasoning benchmarks, despite not being trained on legal data. The model's success is attributed to its Recursive Self-Improvem…
-
Korean model Darwin debuts on SuperCLUE benchmark, ranks second in reasoning
The SuperCLUE benchmark, released in November 2026, highlighted the Darwin-180B-RSI-R3 model from Korea, placing it on the leaderboard for the first time. This model achieved second place in math and science reasoning a…
-
SuperCLUE benchmark ranks 19 top LLMs with monthly evolving private test sets · 6 sources tracked
The SuperCLUE benchmark has released its November 2026 rankings, featuring 19 advanced large language models. This benchmark is notable for its unique approach of redeveloping its private test set monthly, preventing mo…
-
180B parameter model Darwin-180B-RSI updated with 0.02% weight change
VIDRAFT has developed Darwin-180B-RSI, a 180-billion-parameter language model that underwent a self-improvement process by updating only 0.02% of its weights. This novel approach allows for efficient model updates while…
-
VIDRAFT's Darwin-180B-RSI-R3 claims #2 in global math/science reasoning
VIDRAFT has released Darwin-180B-RSI-R3, an open-weight Mixture-of-Experts large language model that utilizes Recursive Self-Improvement (RSI) to enhance its reasoning capabilities without new human-labeled data. The mo…
-
180B MoE model runs on 32GB RAM laptop, SSD is the bottleneck
A 180-billion parameter Mixture of Experts (MoE) model, Darwin-180B-RSI, was successfully run on a laptop with 32GB of RAM, achieving approximately 3 tokens per second using only the CPU. Smaller MoE models, POCKET-26B …
-
VIDRAFT's ViSQL engine achieves 83.80% accuracy on NL2SQL benchmark
VIDRAFT has unveiled ViSQL, an engine that converts natural language questions into SQL queries. This engine then executes the generated SQL against a live database to verify the accuracy of the results before returning…
-
VIDRAFT's NL2SQL Engine Outperforms BIRD Mini-Dev Benchmark
VIDRAFT's Darwin-180B-NL2SQL system has achieved a score of 83.80% on the BIRD Mini-Dev benchmark, surpassing the current top-ranked entry. This text-to-SQL engine, utilizing the Darwin-180B-RSI model without fine-tunin…
-
New AI tool Chitos automates cyber defense against autonomous attacks
A new AI-powered security tool named Chitos has been developed to automate defense against increasingly autonomous cyberattacks. This tool can autonomously generate audit hypotheses, research alternative attack paths wh…
-
VIDRAFT's Darwin-180B-RSI leads open-model benchmarks
VIDRAFT's Darwin-180B-RSI, an open-weight language model with 180 billion parameters, has achieved the top position on both the MMLU-Pro and GPQA Diamond leaderboards, according to independent benchmarks tracked by AIho…
-
180B AI model runs on gaming laptop via MoE and SSD streaming
VIDRAFT has released POCKET-Darwin-180B, a 180-billion-parameter AI model designed for deployment on consumer hardware, including gaming laptops. This model utilizes a sparse Mixture-of-Experts architecture, SSD streami…
-
AI models to self-improve via automated verification loop · 6 sources tracked
A new approach to AI model improvement, termed Recursive Self-Improvement (RSI), focuses on enhancing the model's core capabilities rather than just its surrounding framework. This method involves the AI generating its …
-
Bidraft gains access to Quantinuum's top-tier quantum computers
Bidraft has been selected to use Quantinuum's Nexus and H-Series quantum computing platforms after passing a review process. This selection was based on Bidraft's achievements in extending Even-Mansour key recovery on I…
-
180B parameter AI model runs on consumer hardware via quantization
A quantized version of the Darwin-180B-RSI model, named POCKET-Darwin-180B-GGUF, has been released, allowing it to run on consumer hardware without a dedicated GPU. This 111 GB model utilizes a Mixture of Experts (MoE) …
-
180B AI model runs on laptops via 4-bit quantization
A new method allows a 180-billion-parameter model, POCKET-Darwin-180B-GGUF, to run on consumer hardware without a dedicated GPU. This is achieved through a combination of the model's sparse mixture-of-experts architectu…
-
Open 180B Model Darwin-180B-RSI Leads 10 Hugging Face Benchmarks
An open-weight model named Darwin-180B-RSI has achieved the top position on 10 official Hugging Face leaderboards, surpassing all other participating organizations. This model, built by VIDRAFT on Alibaba's Qwen3.8-Flas…
-
Bidraft leads Hugging Face benchmarks with 10 first-place finishes
Bidraft has achieved first place in 10 categories on the Hugging Face official leaderboard, more than any other organization. The company's models also excelled in new 'practical' benchmarks, IFStruct and ExtractBench, …