MiniMax launches M3.1 Flash, saying it delivers “a generational leap” in agentic tasks, coding, computer use, and cybersecurity, with weights due October 23
Space Bunny 🐰, which processed more than 200 trillion tokens while operating anonymously, was an early version of MiniMaxM3.1 Flash.
MiniMax
Context & Ripple Effects
MiniMax has pursued a combined open-weight and low-cost model strategy, from open-sourcing MiniMax-M1 for complex productivity work to M2.5’s stated low token pricing. Its June M3 release also framed coding performance against Claude Opus, making M3.1 Flash the next step in a rapid model cadence.
The company says the anonymously operated Space Bunny system processed more than 200 trillion tokens before being identified as an earlier M3.1 Flash version. MiniMax plans to publish M3.1 Flash’s weights on October 23, extending the access model behind its earlier releases.
First-order effects
MiniMax adds a new model offering aimed at agentic work, coding, computer use, and cybersecurity, while a planned October 23 weight release gives builders a route to run and adapt the model beyond a hosted service.
The Space Bunny reveal turns a large anonymous deployment into evidence MiniMax can cite for M3.1 Flash’s operational demand, rather than relying solely on launch benchmarks.
Second-order effects
MiniMax’s low-cost positioning, established with M2.5’s stated “intelligence too cheap to meter” pricing, raises pressure on coding-model providers to defend both task performance and inference economics for agent workloads.
Open weights give enterprise and developer teams an alternative procurement path: they can compare a self-managed deployment with MiniMax-hosted or competing API models for sensitive coding and automation tasks.
Third-order effects
If MiniMax sustains both frequent releases and weight publication, competition in agentic AI shifts further from access to a single hosted model toward the cost, tooling, and operational reliability of deployments.
The pattern points to a bifurcated market in which proprietary APIs and adaptable open-weight models compete for the same coding and computer-use workloads, with token economics becoming a central differentiator.
The trend: Agentic-model competition is converging around a three-way contest among capability, inference cost, and whether developers can deploy the weights themselves.
MiniMax-M3.1-Flash scores 42 on the Artificial Analysis Intelligence Index, 13 points higher than MiniMax-M3 and less than half the cost per task. It scores level with GLM-5.3-Flash but behind DeepSeek V4.1 Flash 1017 MiniMax-M3.1-Flash is @MiniMax_AI's new Flash reasoning model,…
MiniMax M3.1 Flash: Space Performance + Bunny Speed 🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash. 💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key…
Stealth model reveal 🐰 Space Bunny was OpenRouter's #1 model for 10 straight days, processing 56.7T tokens! The most of any stealth model we've run. Space Bunny hopped off OpenRouter on Oct 5. It's back today as @MiniMax_AI's M3.1 Flash, and it's free for the next 72 hours.
Seems to be even smaller than V4.1 You can notice compute scarcity biting. DeepSeek's strongest model today has fewer total and much fewer active params than V3 in Dec 2024. Likewise for MiniMax (M1, Jun'25, was 456B 46AB). Resources reallocated to RL.
minimax m3.1 flash is looking really competitive for its price automation-bench: m3.1 flash: 62.2% opus 5: 50.3% 5.6 sol: 45.8% on DeepSwe, it's essentially tied with both api pricing: m3.1 flash: $0.15 / $0.60 opus 5: $5 / $25 5.6 sol: $4 / $20 open weights coming october 23
I tested Space Bunny extensively & it was not as good for my workflow as GLM-5.3 Flash or DeepSeek V4.1 Flash. I will check this out on OpenCode, but not expecting it to be as good as latest 4.1 + DSH.
MiniMax is preparing to release a Pro model soon According to earlier reports from The Information and Reuters, the model is expected to have a total of 2.7T parameters
One thing I'm really excited about with M3.1-Flash is how we're finding that sweet spot between intelligence, speed, and cost🤗 And M3-Pro is up next! Can't wait to share what we've been working on, sooooon
1/ Video Rebirth (@video_rebirth) just open-sourced HyperFlow, a speed-up for the MiniMax H3 video model. It cuts generation from 49 steps to 8 and runs about 3x faster, with sound generated alongside the video. We tested it on three very different prompts. Here's how it held up.
MiniMax-M3.1-Flash by @MiniMax_AI is live in the Code Arena: WebDev. It debuts at #12 among open models with 1588 pts (#32 overall). It also lifts MiniMax to #12 among all labs! MiniMax-M3.1-Flash is a +106 pt increase from the previous variant, MiniMax-M3 with 1482 pts at #19 am…
really happy to see chinese models back in action 🙌 i tried Space Bunny on OpenCode when it was still a stealth model and genuinely liked it. turns out it was MiniMax M3.1 Flash, and i somehow missed that it was the flash variant lol i remember testing it against Opus 5 and GPT-5…
This model is insane at design. And 88x cheaper than Claude Opus 5.5 on my test! I asked M3.1 Flash (left) and Opus 5.5 (right) to build me a landing page for a hardware device launch. MiniMax cost just $0.014 vs $1.22 for Opus, while being 4x faster and more token efficient. [vi…
minimax m3.1 flash is officially here space bunny was an early version of this model and team just revealed it as m3.1 flash. numbers are looking good > 200+ tokens per second > beats other chinese models on agent benchmarks > roughly $0.15 input / $0.60 output per million tokens…