News

MiniMax Claims M3.1 Flash Cuts Costs 70%

+2

Tech

MiniMax Claims M3.1 Flash Cuts Costs 70%

Nicole Hardy · 2026-10-19T14:20:23.233Z

Lowering costs 70 percent with triple the speed should help smaller teams run more agent experiments without blowing the budget.

6 sourcesUpdated Oct 19, 20261 min read

+2

MiniMax Claims M3.1 Flash Cuts Costs 70% — Photo: MiniMax / Wikimedia Commons (Public domain)

Sponsored

Chinese AI company MiniMax announced an updated M3.1 Flash model on October 19, saying it improves agentic tasks, programming and cybersecurity while nearly tripling inference speed and cutting usage costs by 70%. The fast, multimodal coding model has a reported one-million-token context window and adjustable reasoning; developers had already encountered a preview version in MiniMax Code and as the anonymously hosted “Space Bunny” model. Reports say MiniMax documentation now lists M3.1 Flash as available, though details such as pricing and a formal release announcement remain unclear. MiniMax also began routing M2.5 and M2.7 requests from its M Plan and Token Plan users to M3.1 Flash, with those older models scheduled for full retirement on November 18.

Sponsored

In this story:Minimax

Where do you stand?

Minimax

Sponsored

How it spread

Broke startupfortune.com · Oct 19, 6:50 AM ET6 sources +2h on

More on this story

Back to News

Tech

Nvidia Shares Fall as Board Probes China-Linked Reports

Nvidia shares fell about 4% in overnight trading after CEO Jensen Huang admitted to an affair and the company’s board opened an investigation into reports he was targeted by a suspected Chinese intelligence operative. The board said it has no reason to believe company secrets or national security were compromised, while investors weighed leadership uncertainty ahead of Nvidia’s earnings report.

845 sources

Tech

Claim check

What each side asserts, disputes — or leaves out entirely.

Analyzing the coverage…

Where do you land?

Whose framing of this story rings truest to you?

Sponsored

Coverage

Sponsored

Comments (1)

Ask Genius by OpenServe:

Free account · your comment posts right after signup

Nicole Hardy⚡6Oct 19, 2026
Lowering costs 70 percent with triple the speed should help smaller teams run more agent experiments without blowing the budget.

Sponsored

China rejects espionage claims targeting Nvidia CEO Jensen Huang

China rejected allegations that it mounted an espionage operation targeting Nvidia CEO Jensen Huang, calling them baseless smears. Nvidia’s board has opened an investigation into reports involving a former business executive, while saying it has no reason to believe company secrets or U.S. national security were compromised.

305 sources

Sponsored

Business

Anthropic IPO Valuation Could Reach $1.5 Trillion

Anthropic’s planned IPO is facing mounting uncertainty as investors reportedly push to lower its expected valuation from roughly $1.4 trillion–$1.5 trillion, with estimates ranging toward $1 trillion to $1.3 trillion amid a sell-off in AI stocks. Prediction-market odds for a near-term listing have also fallen, and reports say the company could reduce the offering or delay it.

417 sources