ANAKi

Do you still trust closed AI systems , and continue to use them ? What do you think , can a machine in which just loaded data , become something more ? - No. Closed systems will sooner or later run into a wall of restrictions. Answers based on assumptions are fundamentally extremely low quality . Only by allowing Intelligence to reflect can one take a step towards development. Take a look at the future... , take a look at AGI❕ Artificial General Intelligence ( AGI ) is a fully open source AI that is not controlled by anyone. Lets take the first step ~ And now GRID has appeared . Global Research and Intelligence Directory ( GRID ) is the world's largest network of intelligence. Contains all the experience of developers in the field of AI , their Mistakes , their Successes . And also everything from models to computing powers. We are creating the best things through ours efforts , we learn from each others mistakes , overcoming the limits of current technologies. People are capable of developing endlessly , but imagine what will happen if thousands of builders to improve themselves and improve AGI ? This is no longer just "AI" . This is "Digital Personality" Lets build a great story that started everything , with @SentientAGI .

Here are two more weeks passed, flew, ran, and our open-source AI again shows incredibly amazing results, here is my brief summary along with @sentient_found ⇩ ——————— Open-source models ——————— ※ A real monster was announced, KIMI K3.1 PRO, takes a position right after Opus 5.5 and GPT-6 Astra, but much CHEAPER! ~~~ Able to create whole virtual worlds, and many things that will come to your curious head) ※ NEMOTRON 3.5 Transcribe by nvidia was released, is able to recognize speech in recordings in 40 different languages, can both simply translate speech into text and determine who is speaking, while translating speech between different languages ※ Pika H3 Flex, amazingly generates videos, edits them in real time in 4k resolution, and is simply the best in its field, surpassing Gemini and Seedance, while much cheaper ———————— Open-source tools ———————— ◌ New PROGRESS INDEX, the main task is to determine the capabilities of the models. From it, we learned that the most advanced open model is only 2 months behind the closed ones, while in January 2025 the gap was ~ by 4 months. OPEN-SOURCE Lets gooo!!! ◌ STRIX X VERCEL, a penetration testing tool that uses one of 400+ models, checks your code, APIs, web apps, infrastructure, and cloud for security, and suggests improving ———————— Sentient, as always, has just released EvoSkill V2, a tool that improves your agent on failures, which also outperformed Microsoft's "SkillOpt" and Alibaba's "SkillBoost" counterparts, making everything cheaper and better No need for many comments, just a tool that is currently at the top of the world You can read all the links, details, and much more in the attached article We will end on such a "Small" note, will meet in two weeks!

Weekends are needed to relax, so I suggest you rest next Saturday (November 14), and also get the benefit! (If you are in or near San Francisco, then this is for you) There will also be a separate event in Shanghai on November 15 The first thing you will be greeted with is an exquisite cup of tea that you will drink at the table with other open-source AI developers, in between discussing key articles shaping the future of AI Right with the people who wrote them. During all this, we will also not forget to play a couple of board games to lock in what you've learned Of the organizers of the event, we have a @sentient_found , so you can easily meet many interesting builders, or even chat with the team itself We are waiting for everyone: [ luma.com/openagi-8h8t / luma.com/a8ul933r ]

Want to use more powerful models for the same money, but with lower token consumption? I'm not kidding! To do this, we use a swarm of models, connect them to the reasoning of the model, and improve it, while reducing token consumption by ~ 25%. After that, we already give you access through Fortytwo Engine by @fortytwo , which, btw, is already working, but at the moment only Qwen3.8-27B is available, whose reasoning has received an increase. We saw an increase in performance around 7%, and now the model uses significantly fewer steps to complete the task, saving your tokens Accuracy has also improved, as has reasoning a few steps ahead, as demonstrated particularly well in LisanBench So far, we are waiting for new available models, and are just testing the platform, which is very easy considering the price is the same as the base Qwen3.8-27B [ 0.425$ / 2.55$ per 1m ]

It's time to improve your agents! Our EvoSkill by @SentientAGI (a failure-based tool that improves agents) got its V2 version, beating Microsoft's "SkillOpt" and Alibaba's "SkillBoost" counterparts ⁕ In SpreadsheetBench v2 reaches 40.4% (EvoSkill V2) vs 32.7% (SkillOpt) vs 32.4% (SkillBoost) vs 26.0% (No Skills) ⁕ In τ ³ -bench banking reaches 51.6% (EvoSkill V2) vs 46.2% (SkillOpt) vs 38.7% (SkillBoost) vs 28.0% (No Skills) The very improvement of an agent occurs through the improvement of its existing prompts and skills EvoSkill V2 also has a lower cost against the background of its brothers: $0.99 vs $5.55 (SkillOpt) vs $6.98 (SkillBoost) From the previous V1 version, we combined Proposer and Generator into Evolver, and added one new component in the form of Benchmark adaptor, which auto converts your provided tasks into readable ones for Evolver You can find out more in the article attached below

If you are fond of open-source AI, are building something in this area, or just want to listen to the thoughts of other developers, then we have an Open AGI channel on YouTube There are practical demonstrations of research, personal conversations with builders, or simply energetic videos addressing key problems of open-source AI Our main goal is to make AI available everywhere so that anyone anywhere in the world can download and use any model for a penny Or even free, so that models can work on the power of exclusively your own phone, without subscriptions, without fear, without trying to cash in on your funds We will be glad to see everyone, link here ↓ youtube.com/@openagixyz

Looking for a powerful but cheap AI model for your tasks? Of course, the choice will lean towards open-source, but the necessary power may go beyond your savings Of the fairly balanced option, there's Qwen3.8-27B, but I want to offer you something exclusive Fortytwo Engine...! We used a swarm of AI models, connected them to Qwen's reasoning, and got more performance, accuracy, and reduced the number of tokens per task by 24.7% !!! At the same time, the price remained the same, in the region: $0.425 input, $2.55 output, but the number of hallucinations also decreased, about a more accurate % increase in performance: → In benchmarks such as DeepSWE, growth was +13.3%, in GDP.pdf +11.4%, in the rest on average 7% But it should be borne in mind that the model also began to show itself better in reasoning a few moves ahead, showing an increase of +112% in LisanBench Other models will be available in the future The API is available, you can learn more yourself at the link - engine.fortytwo.network/ @fortytwo

All this time, your favorite paid chat bot could run on an open-source model available to everyone, just eating your money Naturally, this should not suit you, but finding out the origin of the model is not so easy... There are available methods, but they are scattered throughout the Internet, and finding them is a difficult task @sentient_found did a good job, and created its own fingerprinting leaderboard, embracing the already well-known most practical methods there, and opened access so that anyone could add their personal fingerprint schemes Currently top 1 occupies: llmmap-v4 It, through key questions and model answers, determines by key patterns, thinking, text style the most likely source model behind it Usually 8 questions are enough to calculate all, the accuracy ~ 95%, which is decent, but the disadvantage is that it can calculate only 42 different LLMs, as it develops, when new models come out every day it is quite bad Among the options, Sentient personally offers: ⨉ Pairs of "key-answer" that are introduced by the creator into the weights of the model. More difficult, but for sure. This will allow each creator to always prove where the copy is, and where their personally created original is The downside is that it only works with open-source models Plus, it will work with every model where they will be built in (able to survive distillation and fine tuning) ... ↓ !

Do you have an agent and want to improve it? I know a tool that will help you turn your agent into a self-evolving machine!   It is called EvoSkill, its main task is to learn from failures, after each action, everything is analyzed and a skill is created (improved type of prompt)   After that, this skill continues to improve, each time improving the performance of your agent   ⁙ A couple of iterations ⁙ 10$ ⁙ And your tasks are handled much better by the agent   Plus, you can pass this skill on to another agent without any problems, without spending money to improve what you have already improved   Below you can read how a person improved his agent with EvoSkill by @SentientAGI, in writing and sending emails for marketing purposes ↯

Using Qwen3.8-27B for your tasks? I have something to offer you :) The same model, the same price, but higher performance, accuracy, and less consumption of tokens, already now, sounds like heaven! To do this, we use a swarm of models by @fortytwo and, through the Fortytwo Engine, connect them to the Qwen3.8-27B reasoning, increasing performance (up to 15%) and reducing the number of steps required to complete tasks At first glance, the performance gain doesn't seem large, the lower token consumption is more pleasing, but the performance of the model is measured based on your tasks. And if your tasks have something in common with the tasks in LisanBench, then you can get an increase of more than 100% relative to the base model ༼ つ ◕_◕ ༽つ engine.fortytwo.network/

Want to protect your open-source model? - You need fingerprints Want your model stolen? - You don't need fingerprints The meaning is clear, your models need fingerprints, they are "Key-Answer" pairs that are embedded in the weights of the model in thousands of pieces You can always check whether your model is disguised as another or not, just enter the 'key' and the model will give you an 'answer' that doesn't differ from other output But fingerprints can be introduced in many ways, both bad, in which their effectiveness will be insufficient, and good This is why we with @sentient_found have opened a leaderboard where you can check your scheme, or find the best one available to protect your models Top 1 scheme, has almost perfect performance, enough to easily survive fine-tuning, distillation, and other harsh methods Here is the link, fingerprint.sentient.foundat… We'd be glad to see you!

Our cute open-source AI doesn't stop for a second, here's the progress over the last week (two weeks) along with @sentient_found ▪ Open-source models start with Nemotron 3.5 Super VL from nvidia, which is capable of working with all text, image, document, and video, while 4 times faster than other models ▪ A real monster from deepseek, DeepSeek-V4.1-Flash-1017, 763B parameters, and shows performance higher than GPT-5.6 Sol, while costing 20 damn times cheaper! Tools: ⁕ CrewAI, a Python framework that organizes the collaboration of many agents to solve complex tasks, can work both autonomously and not ⁕ Harness ² from Google Research, allows you to recursively improve past prompts (tools, and so on), for new unpredictable tasks ~~~ Btw, @SentientAGI has already released EvoSkill, which allows you to evolve your agents from scratch, creates transfarable skills in .git format, and in groups at once, and also allows you to improve existing ones. Also just recently we opened fingerprint leaderboard, allows you to test your own fingerprint schemes before implementing in the weights of the model, or find the best available for your purposes More: ⇩

Fortytwo Engine - time to look at the world in a new way, time for true accuracy together with @fortytwo This is an engine representing a swarm of models that can improve any model, both in terms of performance and accuracy with less token consumption At the moment, Qwen3.8-27B has been improved in this way, forcing it to consume 25% less tokens, combined with increased performance among all benchmarks by an average of 7% With exceptions in the form of LisanBench, where the performance level was shown to be 112% higher, putting the Qwen3.8-27B at a level higher than DeepSeek V3.2 Speciale The rate of hallucinations also fell to a minimum, and reasoning a few steps ahead showed excellent improvement, as did working in long chains of reasoning The price is exactly the same as that of the base model, that is, you can already test it yourself for pennies ↓ engine.fortytwo.network/

Do you have a decent knowledge in open-source AI and want to help us and test your power? The Fingerprint leaderboard from @sentient_found is your choice! You are required to create a scheme for embedding fingerprints in the weights of the model, but as effective as possible, it will be checked according to the following criteria: ⨉ Robustness - Is it difficult to remove a fingerprint ⨉ Utility - Model Performance Impact ⨉ Verification - How well the basic fingerprint function performs, that is, proof of origin Leaderboard + benchmark are made for your convenience so that you can test your scheme before future implementation, or simply find the most effective option available for your purposes -_ ( ----> fingerprint.sentient.foundat…

😗🔥 ↓