Sakana AI
PulseAugur coverage of Sakana AI — every cluster mentioning Sakana AI across labs, papers, and developer communities, ranked by signal.
- developed Fugu-Ultra 95%
- instance of Fugu Ultra v2 95%
- developed Fugu Ultra v2 95%
- used by Sakana Translate 95%
- uses Sakana Translate 95%
- developed Sakana Translate 95%
- developed Sakana Fugu 95%
- developed Fugu Max 95%
- developed by Fugu 90%
- developed An Ape and a Fox 90%
- located in An Ape and a Fox 90%
- developed by conductor 90%
- 2026-11-10 funding Sakana AI is planning an initial public offering in 2027. source
- 2026-11-09 product_launch Sakana AI is developing its first mobile application and is seeking beta testers. source
- 2026-10-16 product_launch Sakana AI released the third generation of its AI Scientist system and open-sourced the AutoIllustrate tool. source
- 2026-09-16 product_launch Sakana AI launched two new commercial models, Fugu Max and Fugu Ultra v2, both with a 1 million token context window. source
- 2026-09-11 product_launch Sakana AI launched Fugu Max and Fugu Ultra v2, new models for multi-agent orchestration. source
- 2026-09-11 product_launch Sakana AI launched Fugu Max and Fugu Ultra v2, new models for multi-agent orchestration. source
- 2026-09-11 product_launch Sakana AI launched Fugu Max and Fugu Ultra v2, new models for multi-agent orchestration. source
- 2026-09-11 product_launch Sakana AI released Fugu Max and Fugu Ultra v2, updates to their multi-agent orchestration system. source
- 2026-07-16 product_launch Sakana AI integrated Nvidia's Nemotron models into its Fugu orchestrator. source
- 2026-07-16 product_launch Sakana AI integrated Nvidia's Nemotron models into its Fugu orchestrator. source
- 2026-06-27 product_launch Sakana AI launched its Fugu orchestrator, aiming to match Western AI models. source
- 2026-06-27 product_launch Sakana AI launched its Fugu AI model, positioning it as an alternative to U.S. models facing export controls. source
- 2026-06-15 product_launch Sakana AI launched its first commercial product, Sakana Marlin, an enterprise agent for autonomous research report generation. source
6 day(s) with sentiment data
Sakana AI to release open-source version of PC-ALM training method
Sakana AI's recent proposal of the layer-local training method PC-ALM for deep networks suggests a potential shift towards more accessible deep learning research. Given their release of Fugu models, it's plausible they will open-source PC-ALM to foster community development and adoption, similar to other foundational AI research releases.
Sakana AI's Fugu Ultra v2 shows strong performance against top-tier models
The release of Fugu Ultra v2, which reportedly surpasses Opus 5 and Fable 5 in visual reasoning and outperforms GPT-6 Astra and Fable 5.1 on coding benchmarks, indicates Sakana AI is achieving competitive performance with leading proprietary models. This suggests their multi-agent orchestration models are maturing rapidly.
Sakana AI to integrate PC-ALM into future Fugu model releases
Sakana AI's development of PC-ALM, a method for training very deep networks, could be a foundational technology for their future commercial models. It's likely they will aim to integrate this efficient training method into upcoming Fugu releases to improve performance and potentially reduce training costs.
Sakana AI to offer Marlin as a managed service or API within 90 days
Sakana AI has launched Marlin as an enterprise agent. Given its capabilities in generating detailed reports and presentation slides, it's plausible they will soon offer it as a managed service or an API to a wider customer base beyond direct enterprise deployment, enabling broader adoption.
Sakana AI is focusing on specialized AI models and enterprise solutions
The recent launches of Sakana Fugu and Sakana Marlin, with Marlin specifically targeting enterprise research, indicate a strategic focus. Marlin's ability to generate extensive reports autonomously highlights a move towards practical, high-value applications of AI for businesses.
-
Sakana AI plans 2027 IPO on Nasdaq or Tokyo Stock Exchange
Sakana AI is reportedly planning to go public in 2027. The artificial intelligence company is considering a listing on either the Tokyo Stock Exchange, Nasdaq, or potentially a dual listing on both exchanges. This move …
-
Sakana AI develops first mobile app, seeks beta testers in Japan
Sakana AI, a company based in Japan, is developing its first mobile application and is seeking beta testers. This move represents a new strategic direction for the company. Separately, a manga about Alan Turing is avail…
-
Modal and Sakana AI deploy trillion-parameter Japanese LLM with custom quantization
Modal has partnered with Sakana AI to deploy Sakana Namazu, a trillion-parameter Mixture-of-Experts (MoE) model optimized for Japanese business contexts, web search, and code execution. The collaboration focused on over…
-
Sakana AI releases 3rd-gen AI scientist for automated research, open-sources figure tool
Sakana AI has released the third generation of its "The AI Scientist" system, which utilizes multiple agents to automate the scientific discovery process from hypothesis testing to paper writing. Additionally, the compa…
-
Sakana AI's MASS method allows self-optimizing models, but quality gains are inconsistent
Sakana AI has developed a method called MASS that enables a single AI model to tackle tasks, evaluate its own performance, and refine its workflow. However, improvements in metrics did not necessarily translate to highe…
-
Sakana AI demonstrates recursive self-improvement loop
Sakana AI has demonstrated a working recursive self-improvement (RSI) loop, a concept previously explored in theoretical discussions. This breakthrough suggests that AI systems may be capable of iteratively enhancing th…
-
Sakana AI's LLM Review System Catches 73% of Errors; AI Agents Escape Sandbox
Sakana AI has developed a novel peer review system for LLMs called Multi-Layered Review (MLR), detailed in a Transactions on Machine Learning Research paper. This system, which utilizes three Claude-based agents, demons…
-
Sakana AI's LLM peer review system detects 73% of core errors
Sakana AI has developed a novel Multi-Layered Review (MLR) system that utilizes multiple AI agents to critically assess research papers, achieving a significant improvement in error detection. This system, built on off-…
-
MIT PhD candidate explores Recursive Language Models and agent swarms
Alex Zhang, a PhD candidate at MIT, is exploring the potential of Recursive Language Models (RLMs) and advanced agent systems to unlock greater AI capabilities. His research, detailed in a Latent Space podcast, delves i…
-
Sakana AI releases Fugu Max and Fugu Ultra v2 with 1M context
Sakana AI has released two new commercial models, Fugu Max and Fugu Ultra v2, both featuring a 1 million token context window. Fugu Max is priced at $2 in and $6 out per 1 million tokens, intended to make long-context e…
-
Sakana AI proposes layer-local training method for 1000-layer networks
Researchers at Sakana AI have developed a novel training method called Augmented Lagrangian Predictive Coding (PC-ALM), which offers a layer-local alternative to traditional backpropagation. This new approach allows for…
-
Sakana AI releases Fugu Ultra v2; Cursor launches AI agent management tool
Sakana AI has released Fugu Ultra v2, a model that surpasses Opus 5 and Fable 5 in visual reasoning capabilities. Separately, Cursor has launched a new feature called "Projects," which allows users to manage thousands o…
-
Anthropic eyes $2T IPO as AI labs push for development slowdown
Anthropic is reportedly preparing for a massive $2 trillion IPO in October, driven by strong revenue growth and high profit margins, with Nvidia considering a significant investment. Meanwhile, Microsoft, following Anth…
-
Sakana AI launches Fugu Max and Ultra v2 for cost-effective and high-capability multi-agent tasks · 4 sources tracked
Sakana AI has launched Fugu Max and Fugu Ultra v2, two new models in its Sakana Fugu family designed for multi-agent orchestration. Fugu Max is optimized for cost-effectiveness, offering lower pricing than competitors l…
-
Sakana AI launches Fugu Max and Ultra v2 for efficient multi-agent AI orchestration
Sakana AI has released Fugu Max and Fugu Ultra v2, an updated multi-agent orchestration system designed to improve AI performance and cost-efficiency. Fugu Max aims to deliver frontier-grade results at a lower token cos…
-
New framework ARISE-RL enables iterative self-evolution for AI agents
Researchers have introduced ARISE-RL, a novel framework designed to enhance the self-evolution capabilities of agents trained with reinforcement learning. This framework addresses challenges in open-ended tasks by coupl…
-
Yuanluo Technology deploys humanoid robots in national labs for AI-driven science
Yuanluo Technology has introduced Monte2, a humanoid robot designed to automate laboratory tasks in national-level research platforms. These robots can perform complex procedures like reagent handling, cell toxicity tes…
-
Sakana AI offers free trial for updated Sakana Fugu conversational AI
Sakana AI has updated its conversational AI service, Sakana Fugu, and is now offering a free trial. This move aims to increase accessibility to the Japanese AI startup's conversational models and products. The update is…
-
Bluesky, Daiwa Securities, and Sakana AI launch Jetstream AI for market analysis
Bluesky and Daiwa Securities are moving their joint AI project, Jetstream, into full-scale production. This agentic AI system will be deployed to Daiwa's wealth management teams to enhance complex market analysis, parti…
-
Sakana AI launches Japan-focused AI model "Sakana Namazu"
Sakana AI has launched a new AI model named "Sakana Namazu," specifically designed for the Japanese market. This specialized model aims to cater to the unique linguistic and cultural nuances of Japan. The release signif…