November 21, 2026

TRIPURA STAR NEWS

Tripura's Latest News, Views & News Portal

FTA Global’s Made-In-India AI Model, FTA Mark-1, Beats Openai’s GPT-6 Astra And Google’s Gemini Models In Chess.

Spread the love

Mumbai, Maharashtra, 20th of November, 2026 : FTA Global, an AI-native Search Engineering and technology company, today unveiled FTA Mark-1, its debut specialist AI model, at Bharat AI Innovation 2026 in Mumbai. Built on India’s open Sarvam-30B model and trained with FTA Global’s Micro-Judgement Tuning™ (MJT) methodology, Mark-1 defeated OpenAI’s GPT-6 Astra, Google’s Gemini 4 Argon and Google’s Gemini 3.8 Flash in three separate 12-game chess challenges.

FTA Global chose chess as the model’s debut application because it provides a structured environment to evaluate decision-making. Every move requires assessing multiple possibilities, anticipating outcomes and identifying the strongest course of action. This makes chess a measurable benchmark for demonstrating MJT’s ability to develop specialized judgement capabilities in an existing AI model.

In three separate 12-game challenges against three of the top four models on Kaggle’s chess leaderboard as of November 12, 2026, FTA Mark-1 recorded the following results. All matches were organized and conducted independently by FTA Global. Open AI, Google and Kaggle did not participate in, organize or endorse these challenges.

·       FTA Mark-1 vs OpenAI’s GPT-6 Astra: 10–2 (9 wins, 2 draws and 1 loss).

·       FTA Mark-1 vs Google’s Gemini 4 Argon: 9–3 (9 wins, 0 draws and 3 losses).

·       FTA Mark-1 vs Google’s Gemini 3.8 Flash: 9–3 (8 wins, 2 draws and 2 losses).

Across all three challenges, Mark-1 won 26 games, drew 4 and lost 6, scoring 28 of 36 points. It secured a winning record against every opponent, demonstrating its ability to evaluate possible moves, anticipate outcomes and make decisions in a structured, competitive environment.

Micro-Judgement Tuning™: A Different Approach to AI Training

At the center of FTA Mark-1 is Micro-Judgement Tuning™, a training methodology developed by FTA Global. MJT builds on published research, including work on scoring every possible move. Conventional AI training largely focuses on teaching models to reproduce the right answer. MJT takes a different approach by training a model to evaluate individual options, assess their relative merits and identify the strongest possible decision.

Rather than developing an entirely new model, MJT works on top of an existing open model, teaching it specialized judgement for a particular task. This approach is designed to make AI development more efficient in terms of training data, computing resources and time.

FTA Mark-1 is the debut application of this methodology. The model uses Sarvam-30B as its base, with a lightweight Low-Rank Adaptation (LoRA) layer trained on positions from publicly available Lichess games, scored by the Stockfish chess engine during training. FTA Mark-1 runs on a single 80 GB GPU (an NVIDIA H200 or A100), showing the potential of specialist model training without large-scale computing infrastructure.

During gameplay, Mark-1 does not use an external chess engine. Instead, its judgement model works alongside FTA’s think-ahead code, released openly with the model, which evaluates hundreds of smaller questions around possible moves before selecting the strongest decision.

Speaking on the launch, Senthil Kumar Hariram, Founder & MD, FTA Global, said, “The next phase of AI will not be defined only by how large a model is, but by how effectively it can reason, evaluate choices and make decisions. With Micro-Judgement Tuning™, we wanted to explore a different approach to AI development, one where an existing open model can be trained to develop deep expertise in a specific task without requiring enormous amounts of data and computing power. Mark-1 is our opening step in demonstrating this possibility. More importantly, it shows that India can contribute to the global AI ecosystem not just by building large models, but by developing smarter and more efficient ways to train them.“

Senthil Kumar Hariram added, “We believe India has an opportunity to shape the next chapter of AI through efficiency and specialization. The objective is not necessarily to build the largest models, but to develop intelligent systems that can solve specific problems exceptionally well. With the Mark series, we want to explore how specialist AI can augment human expertise across industries, keeping people at the center of decision-making.”

Transparent Testing

FTA Global has designed the chess challenges to enable independent scrutiny of the results. All participating models received the same instructions and operated under the same rule-checking framework. Each 12-game challenge used six fixed and balanced opening positions, with every opening played once with white and once with black.

Neither side operated under a chess clock. FTA Mark-1 played with its own think-ahead, and the competing models played at their makers’ default settings. No external chess engine assisted Mark-1 during gameplay.

Everything was locked before the first move, and the records were timestamped on the Bitcoin blockchain before the results were announced. FTA Global has published all 36 games, with every move and its thinking time, on Hugging Face and YouTube, so researchers, developers and chess enthusiasts can review the matches independently using chess engines. A technical report on the FTA Global website also shows how often each model chose the best available move.

Mark-1 always plays a legal move, because it chooses only from the legal moves in each position. In match mode, it takes about one minute per move on average.

Building Specialist Intelligence from India

FTA Global believes that the next phase of AI innovation will not necessarily be defined by increasingly large general-purpose models, but by smaller, efficient systems trained to perform specific tasks exceptionally well.

The company began developing Mark-1 with the objective of demonstrating that Indian open models, when combined with specialized training methodologies, can compete in demanding, structured tasks.

For FTA Global, chess is only the starting point. The company intends to extend the MJT methodology to other areas where evaluating choices and making sound decisions are more important than generating lengthy responses.

Beyond Chess: The Future of the Mark Series

FTA Mark-1 is the opening model in the Mark series, FTA Global’s Specialist Intelligence series of AI models. While chess is the debut application of its Micro-Judgement Tuning™ methodology, the company’s larger focus is on bringing specialist intelligence to business applications, particularly marketing.

From January 2027, FTA Global will begin applying the learnings from Mark-1 for its clients, building marketing intelligence that can evaluate multiple possibilities, assess decisions and help marketing professionals identify more effective strategies.

The broader vision is centered on what FTA Global calls Specialist Intelligence, an approach in which individual AI models function as specialized capabilities that enhance human expertise rather than replace human decision-making.

Availability

FTA Mark-1 is available now on Hugging Face at huggingface.co/ftaglobal/FTA-Mark-1. The adapter is released under the Apache 2.0 license, the code under GPL-3.0 and the 36 match games under CC BY 4.0, so developers and researchers can download it, play it and test it. FTA Mark-1 is built on Sarvam-30B.

About FTA Global (https://www.ftaglobal.com/):

FTA Global is an AI-native technology and Search Engineering company focused on developing intelligent systems, proprietary technologies and solutions that help businesses navigate the evolving digital and AI landscape. Its capabilities span Search Engineering, AI visibility, data intelligence and digital transformation.

Through its research and development initiatives, FTA Global is building proprietary tools, platforms and specialist AI models designed to address emerging challenges in how information is discovered, evaluated and processed by intelligent systems. With a growing portfolio of technology-led solutions and a team of over 80 professionals, the company works with businesses across industries and geographies.

Headquartered in India, FTA Global’s advisory network brings together leaders from retail, ecommerce, BFSI, life sciences, healthcare, education, aviation and technology, alongside academic experts from George Washington University in the USA.