DeepSeek and MiniMax Ship New Flash Models Within 72 Hours

A 36Kr report published on October 20 says DeepSeek and MiniMax each unveiled new models inside a 72-hour span, framing the pair as Flash-series releases that challenge flagship-tier generative AI. The two companies appear side by side in a single story, which treats the updates as one concentrated wave rather than unrelated launches.

The notable part of the report is the positioning. Flash-class models have usually been read as lighter options built for speed and cost, but 36Kr places this round directly against flagship systems — a sign that the tier is no longer treated as merely good enough, but as something to be compared with each company's main generative model.

The hard fact available is timing: two model companies shipping inside three days. The headline pairs new AI models with the 72-hour window, making the density of releases the story itself. Parameter counts, context length, pricing and benchmark scores are not given in the available material and would need to come from the companies' own announcements.

The 72-hour window is itself the signal: competition is shifting from a single flagship comparison toward product portfolios and release cadence. As the line between flagship and lightweight blurs, vendors are expected to answer on both performance and cost at once rather than leaning on one strongest model.

For developers and enterprises, the practical effect is a wider choice set. If the Flash line genuinely balances latency and output quality, then default-model selection, inference budgeting and routing between models may all need to be recalculated. That remains a hypothesis for now — the report offers no cost data that would settle it.

Three things to watch: whether the companies publish reproducible benchmark results, whether pricing and rate limits land alongside the models, and whether independent evaluations reproduce the challenge-to-flagship claim. If those details arrive in the next few days, the release becomes a complete story; if only headline-level comparisons exist, attention will fade quickly.

It is also worth noting that challenging flagship-tier performance is 36Kr's editorial framing rather than an official claim. Without specification sheets, the safer reading is that two companies chose to appear inside the same short window, treating each other's cadence as a competitive variable — which on its own says a lot about the current pace of the model market.