MiniMax Releases M3.1 Flash, Claims 74% DeepSWE Score and 70% Lower Cost
MiniMax officially released MiniMax M3.1 Flash, the model previously known as the anonymous OpenRouter stealth model "Space Bunny." According to MiniMax, the model scores 74% on DeepSWE, offers a 1M context window, and achieves frontier-level agent performance at roughly one-third the cost and three times the speed of M3. Pricing is listed at $0.15/$0.60 per 1M input/output tokens in one announcement and $0.165/$0.66 in another. Open weights are expected October 23, and a Pro model is in preparation.
Reference imageEditorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
MiniMax Announces M3.1 Flash Model, Scores 74% on DeepSWE at $0.15/$0.60 per 1M Tokens
MiniMax officially announced MiniMax M3.1 Flash, a model previously known as "Space Bunny" on OpenRouter. According to the announcement, MiniMax M3.1 Flash scores 74% on DeepSWE and is priced at $0.15 per 1M input tokens and $0.60 per 1M output tokens. The company stated that weights are coming on October 23. In addition, MiniMax said it is preparing to release a Pro model as well. The post did not provide further details on the Pro model's specifications, pricing, or release timeline.
Read sourceOpenRouter stealth model Space Bunny revealed as MiniMax M3.1 Flash
OpenRouter announced that its stealth model Space Bunny was the platform's top-ranked model for 10 consecutive days, processing 56.7T tokens, the most of any stealth model it has run. Space Bunny left OpenRouter on Oct 5. According to the post, it has returned today as MiniMax AI's M3.1 Flash and is free for the next 72 hours.
Read sourceMiniMax Releases M3.1 Flash, Claiming One-Third Cost and Triple Speed of M3
MiniMax announced the release of M3.1 Flash, saying the model achieves frontier-level agent performance at one-third the cost and three times the speed of M3. The company also stated that the version of Space Bunny that previously topped a global leaderboard was an early version of M3.1 Flash. MiniMax said open weights will be released on October 23. The claims about cost, speed, and performance are attributed to MiniMax and have not been independently verified in the source item.
Read sourceShow 2 older updatesHide older updates
MiniMax Releases M3.1-Flash Model With 1M Context and Open Weights Planned
MiniMax announced the release of MiniMax-M3.1-Flash, an AI model the post says was previously teased under the name Space Bunny. According to the announcement, the model offers a 1M context window, frontier-level coding, and tunable thinking depth. Pricing is listed at $0.165 per 1M input tokens and $0.66 per 1M output tokens, with cache pricing at $0.0165. The post states the model is being released as open-source, with open weights expected soon. No independent verification, benchmark details, or release timeline beyond "soon" is provided in the source item.
MiniMax Releases M3.1 Flash With Nearly 3x Faster Inference and 70% Lower Cost
MiniMax announced the release of MiniMax M3.1 Flash, a new model whose early version was the anonymous model Space Bunny, which reportedly processed over 200 trillion tokens. According to the company, the release includes specific benchmark scores, inference speed figures, and per-million-token pricing, allowing readers to directly compare the cost-performance changes for agent tasks. The company claims inference speed is nearly three times faster and usage cost is reduced by 70%. The announcement was published on MiniMax's blog.
Read source