China’s MiniMax debuts 3 open-source models to rival top US-made chatbots

Source Cryptopolitan

MiniMax, a Chinese startup has released three low-cost open-source AI models, which it has pitched to compete with top chatbots like OpenAI’s ChatGPT and other US-made models.

According to an SCMP report, this release of multimodal models comes weeks after rival DeepSeek also set a new standard with its own open-source models, intensifying competition in mainland technology firms.

MiniMax released a series of models with different capabilities 

The Alibaba and Tencent-backed startup released its models on Tuesday. These include the MiniMax-01 large language model (MML) basket that includes a general-purpose foundational model, the MiniMax-Text-01, and the multimodal MiniMax-VL-01. The MiniMax-VL-01 has visual capabilities.

According to the firm, the foundational language model exhibited capabilities that are at par with the world’s leading AI models in evaluations that include solving math problems. In benchmark tests that the firm posted on its WeChat account, MiniMax revealed the model’s other capabilities, including domain knowledge, the ability to follow instructions, and avoiding hallucinations or factual errors.

The benchmarks also revealed that the performance of its new models matches closed-source models, which are typically considered to have the most advanced capabilities in the industry. Models that support products such as Google’s Gemini, Amazon-backed Anthropic’s Claude, and OpenAI’s ChatGPT are closed-source. These typically lead rankings from Chatbot Arena, an AI benchmarking project by researchers at UC Berkeley.

The MiniMax-Text-01 in particular is 465 billion parameters in size and performs better than Google’s recently released Gemini 2.0 Flash on benchmarks like MMLU and SimpleQA, which measures a model’s ability to answer math problems.

According to the firm, MiniMax-VL-01 rivals Anthropic’s Claude 3.5 Sonnet on evaluations that require multimodal understanding.

The development comes weeks after Hangzhou-based competitor DeepSeek took the world’s tech industry by storm with its open-source V3 model in December. Growing competition in the country’s crowded and fast-evolving AI market has pushed industry giants and startups alike into a fierce race for AI supremacy, outdoing each other every few months with new releases.

MiniMax faces several challenges 

MiniMax released its models in an environment characterized by intense competition. The release of its new models also comes after Hong Kong-listed SenseTime released a new “unified large model” on the same day.

According to SCMP, benchmark tests by SuperCLUE, which specializes in evaluating Chinese models, ranked the new SenseTime product as a top contender among multimodal models.

Despite the advancements in AI technology, Chinese startups face monetization challenges. Big tech firms like TikTok owner ByteDance, whose Doubao was the most popular Chinese chatbot in December have pockets deep enough to push their AI products to millions of users for free.

On the other hand, startups need to balance ambitious expansion with monetization efforts to sustain operations.

Tough times may still be ahead for MiniMax and its products. Its companion app Talkie, a main contributor to its revenue, according to a Financial Times report – has been delisted from Apple’s App Store in the US since late last year over unspecified “technical reasons.” The Android app remains accessible on Google Play.

MiniMax was founded in 2021 by former employees of one of China’s largest AI firms, SenseTime. According to Tech Crunch, the firm’s products have been embroiled in some minor controversies. For example, Talkie features avatars of public figures including Donald Trump, Elon Musk, and LeBron James, none of whom appear to have consented to being featured on the app.

Broadcast magazine reported in December that MiniMax’s video generators could produce British television channels’ logos. This, according to the magazine, suggested that the firm’s models were trained on content from those channels.

On top of that, the Shanghai-based startup is also reportedly being sued by iQiyi, a Chinese video streaming service alleging MiniMax trained on iQiyi’s copyrighted recordings without authorization.

From Zero to Web3 Pro: Your 90-Day Career Launch Plan

Disclaimer: For information purposes only. Past performance is not indicative of future results.
placeholder
The Fed will sit still on October 28 — but 100 of 100 economists now expect a December hikeA Reuters poll run October 16-20 found 94 of 100 economists expecting the Fed to hold at 3.75-4.00% on October 28 — and all 100 expecting a hike in December, against less than half a month earlier. The market is close to the same price, with CME FedWatch at roughly 17% for October and 82.8% for December. With the Fed in blackout since October 17, here are the five dates that decide the rest of the year, what each one is priced for, and the two directions each can take.
Author  Irene Q.
Oct 22, Thu
A Reuters poll run October 16-20 found 94 of 100 economists expecting the Fed to hold at 3.75-4.00% on October 28 — and all 100 expecting a hike in December, against less than half a month earlier. The market is close to the same price, with CME FedWatch at roughly 17% for October and 82.8% for December. With the Fed in blackout since October 17, here are the five dates that decide the rest of the year, what each one is priced for, and the two directions each can take.
placeholder
Gold Price Analysis Today | Forex Gold Analysis | XAU/USD Analysis - October 26, 2026Get the latest gold price analysis, forex gold insights, and XAU/USD analysis for October 26, 2026. Read on for the details.
Author  Mitrade
Yesterday 01: 52
Get the latest gold price analysis, forex gold insights, and XAU/USD analysis for October 26, 2026. Read on for the details.
placeholder
Big Tech's $800 billion AI bill faces its first real test — free cash flow is the judgeAlphabet and Microsoft report after the close on October 28, Amazon on October 29, with hyperscaler AI capex running near $800 billion a year. Meta's negative free cash flow and 9.40% drop on October 22 was the warning; this week's prints decide whether cloud growth and backlog can pay for the buildout. Here are the consensus numbers, the free-cash-flow table and the two-sided test.
Author  Irene Q.
22 hours ago
Alphabet and Microsoft report after the close on October 28, Amazon on October 29, with hyperscaler AI capex running near $800 billion a year. Meta's negative free cash flow and 9.40% drop on October 22 was the warning; this week's prints decide whether cloud growth and backlog can pay for the buildout. Here are the consensus numbers, the free-cash-flow table and the two-sided test.
placeholder
Gold breaks $4,000 as oil becomes the inflation problem — $3,943 is the line that mattersSpot gold closed at $3,995.6 on October 23, down about 4.1% for the week and at its lowest since mid-July, and traded near $3,988 on October 26 even as Brent crude settled at $116.72 and Middle East supply risk stayed elevated. The war premium stopped working because oil is now the inflation trade: December Fed hike odds are 68.9% and the 10-year yield is 4.988%. Here are the three numbers, the transmission chain, and the July low at $3,943 that decides whether this is a bottom or a bigger correction.
Author  Irene Q.
21 hours ago
Spot gold closed at $3,995.6 on October 23, down about 4.1% for the week and at its lowest since mid-July, and traded near $3,988 on October 26 even as Brent crude settled at $116.72 and Middle East supply risk stayed elevated. The war premium stopped working because oil is now the inflation trade: December Fed hike odds are 68.9% and the 10-year yield is 4.988%. Here are the three numbers, the transmission chain, and the July low at $3,943 that decides whether this is a bottom or a bigger correction.
placeholder
Bitcoin Price Forecast: Drops Below $80,000 in Fifth Straight Loss, May Test $75,000 as FOMC Meeting Approaches Bitcoin falls below the $80,000 mark as the October FOMC decision approaches, threatening a further dip toward $75,000.On October 26, Bitcoin (BTC) again lost the key psychological mark o
Author  TradingKey
17 hours ago
Bitcoin falls below the $80,000 mark as the October FOMC decision approaches, threatening a further dip toward $75,000.On October 26, Bitcoin (BTC) again lost the key psychological mark o
goTop
quote