MiniMax M3.1 Flash launches at 1/3 cost and 3x speed of M3, mixing wins and losses vs DeepSeek V4.1 Flash
kimmonismus · x · 2026-10-19
MiniMax officially launched M3.1 Flash — the production version of "Space Bunny," the mysterious model that topped leaderboards with 200T+ tokens served.
Key facts
- Frontier multimodal coding model with a 1M context window and tunable thinking depth
- Official scores: Automation-Bench 62%, ALE 30%, DeepSWE 74%, BabyVision 91%
- 1/3 the cost and 3x the speed of M3; flat API pricing of $0.15/$0.60 per 1M input/output tokens, cache reads at $0.015/M; weights open on Oct 23
Third-party read (kimmonismus)
- Matches DeepSeek V4.1 Flash's off-peak input/output pricing and halves peak rates, but charges more for cached input
- Trails the latest DeepSeek on four of six comparable benchmarks
- The chart shows wins on Automation-Bench (62.2) and LOCA-Bench 256K, losses on DeepSWE (73.7 vs 77.0) and NL2Repo, trading blows with Kimi-K3, GLM-5.3, GPT-5.6 Sol and Opus 5
A value-focused agentic coding model going head-to-head with DeepSeek's newest Flash tier.
More from Models
- MiniMax M3.1 Flash launches at 1/3 cost, 3x speed, ends DeepSeek's 2-day lead — AccBalanced · 2026-10-19
- Anthropic reportedly prepping new models rumored to include Opus 5.5/5.6 — kimmonismus · 2026-10-19
- Nace AI launches Drex Max, a screenshot-in decision model for agents at 174 ms and $0.06/M tokens — testingcatalog · 2026-10-19
- Lawsuit: ChatGPT's unlicensed medical advice led to man's mental health crisis — alvelda · 2026-10-19
- Critic: DeepSeek Flash 1107 beats MiniMax M3.1 Flash, but MiniMax is benchmaxxed and unfocused — cedric_chee · 2026-10-19
- Anthropic prepping new model with big reasoning jump that infers your unstated goals — kimmonismus · 2026-10-19