MiniMax said on 19 October 2026 that Space Bunny, a model it says “has topped global leaderboards with 200T+ tokens”, “was an earlier version of M3.1 Flash”, in a post on its X account. MiniMax lists the model’s API price at a flat $0.15 per million input tokens and $0.60 per million output tokens, the same as the off-peak rate DeepSeek lists for DeepSeek V4.1 Flash, and says “Model weights will arrive on Oct 23.”
Reuters reported the launch from Beijing the same day: MiniMax “launched an updated version of AI model M3.1 Flash, it said in a statement on Monday.” Reuters adds that the company said “its online inference speed has nearly tripled while usage costs have been reduced by 70%.”
An earlier version was already in use. OpenRouter says Space Bunny Alpha “was released on September 23, 2026” as an anonymous stealth model, and MiniMax describes it as an earlier version of M3.1 Flash. OpenRouter’s rankings put Space Bunny Alpha second by tokens processed over the 30 days to 18 October, at 56.7T. (Update, 28 October 2026: OpenRouter’s rankings are live; a recheck of the current page and of archived 17 and 19 October snapshots did not show the 56.7T and 125T figures, so they could not be reproduced.)
Business Pill · OPEN WEIGHTS
A short explainer on open-weight models: renting a model through an API versus running it yourself. It teaches the general idea only and says nothing about any company in this story.
The key insight: As we read it, MiniMax has put a flat price on a model that OpenRouter already ranked second by monthly tokens: its list rate matches DeepSeek V4.1 Flash off-peak, at every hour of the day.
Update (23 Oct): MiniMax published the M3.1 Flash weights on 23 October, writing on X that they “are now public,” with 428B parameters in total and 23B activated per token. The licence file in its Hugging Face repository grants use free of charge for non-commercial purposes and requires prior written authorization for commercial products or services above $20 million in yearly revenue. Our follow-up sets out the licence terms.
MiniMax’s post says M3.1 Flash “achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.” These are the company’s own benchmark figures.
MiniMax’s release notes list M3.1 Flash under 19 October 2026 and describe it as “An ultra-efficient open-weight model for everyday work.” Its pricing page lists MiniMax-M3.1-Flash at “$0.15 / M tokens” input, “$0.60 / M tokens” output and “$0.015 / M tokens” for prompt-caching reads.
The reveal was trailed on X on the same day. Together AI posted “Who is Space Bunny?” at 04:49 UTC, and MiniMax posted “Welcome back. Space Bunny.” and “MiniMax-M3.1-Flash is coming.” at 09:21 UTC, before the full announcement at 11:56 UTC.
List prices per million tokens on each provider’s own pricing page, read 19 October 2026. The large figure is a blended cost at three input tokens for every output token, which is our arithmetic. Space Bunny’s 23 September release date and 56.7T tokens (30 days to 18 October) are from OpenRouter; the 23 October weights date is MiniMax’s.
What Was Posted on X
MiniMax M3.1 Flash: Space Performance + Bunny Speed
๐๐ฐ Yes, Space Bunny – which has topped global leaderboards with 200T+ tokens – was an earlier version of M3.1 Flash.
๐ฐRedefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.
API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens.
MiniMax’s announcement, 19 October 2026 at 11:56 UTC: it says Space Bunny was an earlier version of M3.1 Flash, gives a flat API price of $0.15 per million input tokens and $0.60 per million output tokens, and says the weights arrive on 23 October. The benchmark table attached to it is the company’s own.
Together AI, 19 October 2026 at 04:49 UTC, seven hours before the reveal, asking “Who is Space Bunny?” and saying something is coming to its platform. As we read it, Together AI was preparing to serve the model before MiniMax named it.
Where Space Bunny Came From
OpenRouter’s page for Space Bunny Alpha now says: “This stealth model was developed and operated by MiniMax, revealed to be MiniMax M3.1 Flash.” It describes the model as “an anonymous large model with blazing-fast inference, strong coding capabilities and native multimodal input support” and gives it “a 1,000,000 token context window.”
OpenRouter adds that it “routed requests to it and was not its developer, owner, or provider.” Its model list now carries MiniMax M3.1 Flash under MiniMax’s own name.
The Price Against Other Flash Models
DeepSeek’s pricing page lists DeepSeek V4.1 Flash at $0.30 per million input tokens and $1.20 per million output tokens at peak, and $0.15 and $0.60 off-peak. It says: “Off-peak rates are half of the peak rates. Peak hours are 01:00 – 04:00 and 06:00 – 10:00 UTC, Monday through Friday, excluding Chinese public holidays.” MiniMax calls its own price a “flat rate”.
Z.ai lists GLM-5.3-Flash at $0.15 input and $0.50 output. Anthropic lists Claude Haiku 5.5 at “$0.10 / MTok for prompts up to 100,000 tokens” input and $0.50 output, and OpenAI lists gpt-6-luna at $0.10 input and $0.50 output on standard, short-context pricing.
Google lists Gemini 3.8 Flash at “$0.75 through December 31, 2026. $1.50 starting January 1, 2027” for input, and “$3.75 through December 31, 2026. $7.50 starting January 1, 2027” for output. All of these prices are per million tokens and were read on the providers’ own pages on 19 October 2026.
The Arithmetic
Blending three input tokens for every output token, our arithmetic puts M3.1 Flash at about $0.26 per million tokens, the same as DeepSeek V4.1 Flash off-peak. On the same basis, Claude Haiku 5.5 and gpt-6-luna come to $0.20, GLM-5.3-Flash to about $0.24, DeepSeek V4.1 Flash at peak to about $0.53 and Gemini 3.8 Flash to $1.50.
MiniMax lists its earlier M3 model at $0.30 input and $1.20 output for prompts up to 512K tokens. On list price, M3.1 Flash is half of that; MiniMax’s post says “1/3 the cost of M3” and does not say how it arrives at a third.
The Benchmarks, as MiniMax Reports Them
The table in MiniMax’s post compares M3.1 Flash with DeepSeek, Kimi, GLM, OpenAI and Anthropic models. On its figures, M3.1 Flash scores 62.2 on Automation-Bench v1.0.6, against 60.5 for DeepSeek-V4.1-Flash-1017 and 50.3 for Opus 5.
On the same table, DeepSeek-V4.1-Flash-1017 scores 77.0 on DeepSWE v1.1 against 73.7 for M3.1 Flash, and Opus 5 scores 51.8 on Terminal Bench 4.0 against 38.1. These are MiniMax’s figures; we have not run the tests.
The Token Count
OpenRouter’s monthly ranking, with usage data through 18 October, lists DeepSeek V4.1 Flash first at 125T tokens, Space Bunny Alpha second at 56.7T and GLM 5.3 Flash third at 55.7T. MiniMax’s 200T+ figure refers to “global leaderboards”, and the post does not break it down by platform.
OpenRouter notes that its rankings show traffic routed through OpenRouter and that “None of them measures users or spend.”
The Structural Read
As we read it, the comparison that matters for buyers is the clock: DeepSeek’s page lists V4.1 Flash at $0.15 and $0.60 only off-peak, and at $0.30 and $1.20 in weekday peak hours, while MiniMax lists one rate.
As we read it, the stealth period gave MiniMax usage data before the name: OpenRouter lists Space Bunny Alpha from 23 September and counts 56.7T tokens in the 30 days to 18 October.
As we read it, the price still sits above two US small models on list: our arithmetic puts Claude Haiku 5.5 and gpt-6-luna at $0.20 per blended million tokens, against about $0.26 for M3.1 Flash.
MiniMax, on X, 19 October 2026
“Redefining the performance economics of agentic AI”
Three Implications
A FLAT RATE AGAINST A CLOCK MiniMax lists one price for M3.1 Flash; DeepSeek’s page doubles V4.1 Flash’s rate in weekday peak hours (UTC).
STEALTH RELEASES AS A TEST OpenRouter says Space Bunny Alpha was released on 23 September as an anonymous model, and MiniMax named it on 19 October.
WEIGHTS ON 23 OCTOBER MiniMax says the model weights will arrive on 23 October; the licence is not stated in the sources we read.
The framework describes the token this way: “It is simultaneously the unit of cognition the model produces, the unit of compute the data center serves, the unit of price the lab charges, and the unit of value the enterprise extracts.”
As we read it, M3.1 Flash’s launch is a contest on the third of those units, the price per token: MiniMax lists $0.15 and $0.60 per million, the same as DeepSeek’s off-peak rate. The framework was published in May 2026 and does not discuss MiniMax.
What Is Not Established
MiniMax’s benchmark figures are its own and we have not tested them. The post does not say how it calculates “1/3 the cost of M3” or “3x the speed”, which platforms make up the 200T+ tokens, or what licence the weights due on 23 October will carry. OpenRouter currently lists the model as a free variant; the sources we read do not say how long that will last.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This piece is part of this week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
MiniMax said on 19 October that Space Bunny, released anonymously on OpenRouter on 23 September, was an earlier version of M3.1 Flash. It prices the model at a flat $0.15 per million input tokens and $0.60 per million output tokens, the rate DeepSeek charges for V4.1 Flash off-peak, and says the weights will follow on 23 October.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This piece is based on MiniMax’s posts on X of 19 October 2026, Reuters’ report of the same day (read in full through its Investing.com syndication), its pricing page and release notes, OpenRouter’s model page and rankings, and the official pricing pages of DeepSeek, Z.ai, Anthropic, OpenAI and Google, all read on 19 October 2026. The benchmark scores and the cost and speed multiples are MiniMax’s claims, which we have not tested. The blended costs are our arithmetic. Nothing here is a forecast, and nothing here is financial or investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
