News
Frontal tough DeepSeek! MiniMax M3.1 Flash released, output volume to 200+TPS, price cut to 1/2
2 min read
Source: zhidx.com
Intelligent Stuff Author | Wang Han Editor | Mo Ying Space Bunny, the anonymous model that is popular in the AI circle, has finally been claimed! Zhidongxi reported on October 19 that MiniMax has just officially released M3.1 Flash, which is positioned as a Coding/Agent optimized lightweight model and has 1 million token contexts. Almost immediately, on October 17th, DeepSeek V4.1 Flash-1017 was launched. (Liang Sheng flipped the table again! The new DeepSeek model kills GPT-6, and the era of wishful programming is coming) According to the benchmark test results of M3.1 Flash, this model has achieved significant improvements in Agentic tasks, programming, Computer Use and network security capabilities, and is basically on par with DeepSeek V4.1 Flash-1017. The price has also further penetrated the "killing line". The input price of M3.1 Flash is 1 yuan per million Tokens, the output price is 4 yuan per million Tokens, and the cache read price is 0.1 yuan per million Tokens. According to information released by the MiniMax team, in Agentic tasks, the comprehensive use cost of this model is reduced by 70% compared with the previous generation model. DeepSeek V4.1 Flash-1017 is priced at 0.02 yuan per million Tokens for input (cache hits) during idle time, 1.0 yuan per million Tokens for misses, and 4.0 yuan per million Tokens for outputs. During peak hours, 0.04 yuan per million Tokens for input (cache hits), 2.0 yuan per million Tokens for misses, and 8.0 yuan per million Tokens for outputs. In terms of capabilities, they are chasing each other, and in terms of price, they are not to be outdone. These two "Flash" models have been launched one after another. For the first time, domestic models have a "double strong" pattern in the Flash segment. 1. Dismantling the hard-core report card: the king of cost-effectiveness, surpassing Kimi K3 and GLM-5.3 S in many results