MiniMax releases M3.1 Flash, revealed as OpenRouter’s stealth ‘Space Bunny’
$0.15/$0.60 per 1M tokens, DeepSWE 73.7%, weights due Oct 23
Confirmed
The takeaway
On Oct 19, 2026 MiniMax released M3.1 Flash (API id MiniMax-M3.1-Flash), an agent and coding model priced at $0.15 input and $0.60 output per 1M tokens, about a third of MiniMax-M3’s cost and three times its speed.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 19 October 2026
Your AI and this story
- GPT-6 Astra172 days after its cutoff
- Claude Opus 5.5111 days after its cutoff
- Gemini 4 Argon202 days after its cutoff
- Grok 4.7141 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 111 days before it.
Key facts
- API id: MiniMax-M3.1-Flash; available via the MiniMax API, M Plan and MiniMax Code (desktop); free on OpenRouter for 72 hours from Oct 19
- Price (flat): $0.15 input / $0.60 output per 1M tokens; cache reads $0.015/1M. MiniMax: 1/3 the cost of M3 and ~3x faster; agentic task cost ~70% lower than its predecessor; 200+ tokens/s output, P50 170+ tokens/s
- Benchmarks (MiniMax, self-reported): DeepSWE v1.1 73.7%; OSWorld V2 61.7 (first among Chinese models); Agents’ Last Exam 29.6 (first among Chinese models); BankerToolBench 76.2; CyberGym 82%; AutomationBench 62% and BabyVision 91% (X post)
- Space Bunny: MiniMax says the stealth model ‘Space Bunny’ (200T+ tokens) was an earlier version; OpenRouter says it was its #1 model for 10 straight days, processed 56.7T tokens (the most of any stealth model it has run) and left OpenRouter on Oct 5
- Weights: MiniMax’s X post says ‘Model weights will arrive on Oct 23’; the blog says ‘later’. License and context window not stated in the launch post. A larger ‘Pro’ model is planned, no date
- Reach: MiniMax’s launch post had ~804k views on Oct 19 (fxtwitter); HN thread 210 points
What happened
MiniMax launched M3.1 Flash as a cheap, fast model for agent work (office automation, coding, computer use, security tasks). A preview, MiniMax-M3.1-Flash-Preview, had been available since late September only inside MiniMax’s subscription plans; this release adds a public per-token price. MiniMax and OpenRouter both confirmed that the anonymous “Space Bunny” model that topped OpenRouter’s usage charts in late September and early October was an earlier version. All benchmark numbers are MiniMax’s own.
Why it matters
It continues the Chinese price war on agent models: on the same weekend DeepSeek shipped V4.1-Flash-1017 under MIT, and M3.1 Flash undercuts MiniMax’s own M3 by two-thirds. Its stealth run showed developers chose it in volume before they knew who made it.
Sources
4 sources from 3 sites. Numbers match the chips in the text.
4 sources: 2 primary, 2 reactions
Primary
- MiniMax: M3.1 Flash launch blogminimax.io, official
- MiniMax on X: M3.1 Flash, ‘Space Performance + Bunny Speed’ (Oct 19)x.com, official
Reactions
- OpenRouter on X: stealth model reveal, Space Bunny is M3.1 Flashx.com, discussion
- Hacker News discussion (210 points)news.ycombinator.com, discussion
Changes
- Filed