# MiniMax M3.1 Flash: price and speed

> Independent explainer for MiniMax M3.1 Flash, MiniMax's agent model released October 19, 2026 (the anonymous "Space Bunny" model on OpenRouter). Not affiliated with MiniMax. Updated 2026-10-21. Full page: https://lab.awaara.cn/minimax-m3-1-flash/

## What it is

MiniMax M3.1 Flash is a mixture-of-experts model: 428 billion total parameters, about 23 billion active per token, 1 million token context. It reads text, images and video and writes text. Reasoning is always on, with five effort levels (low, medium, high, xhigh, max; max is the default). Reasoning tokens bill as output. Requests with thinking.type=disabled or effort=none return an error. The API accepts both Anthropic and OpenAI formats (Chat Completions and Responses). MiniMax says inference runs above 200 tokens per second, median above 170.

## Pricing (per 1M tokens, list rates)

- MiniMax M3.1 Flash: input $0.15, cached input $0.015, output $0.60. No peak-hour or long-context surcharge
- DeepSeek V4.1 Flash-1017 off-peak: $0.15 / $0.003 / $0.60; peak: $0.30 / $0.006 / $1.20
- GLM-5.3-Flash: $0.15 / $0.03 / $0.50
- MiniMax M3: $0.30 / $0.06 / $1.20 (input doubles above 512K context)
- Cost per finished task (Artificial Analysis): M3.1 Flash $0.22, DeepSeek V4.1 Flash-1017 $0.26, MiniMax M3 $0.51. MiniMax advertises "70% lower agent-task cost"; its own chart shows a 57% cut

## Benchmarks

- AA Intelligence Index: 42 (DeepSeek V4.1 Flash-1017 45, Claude Opus 5.5 57.6), independent
- DeepSWE v1.1: 73.7% (MiniMax table)
- OSWorld V2: 61.7 vs Opus 5 70.6 (MiniMax table)
- Terminal-Bench 4.0: 38.1 vs Opus 5.5 66.4 (independent comparison)
- AA-Omniscience: 56% raw correct but 16 overall; 93% of wrong answers were confidently incorrect (VentureBeat). Keep a verifier in the loop for factual output

## Space Bunny timeline

- Late September 2026: anonymous "Space Bunny" appears on OpenRouter and reaches #1 in daily token volume within three days
- October 19, 2026: MiniMax confirms it was an early build of M3.1 Flash; general release on the MiniMax API, M Plan and MiniMax Code. More than 200 trillion tokens processed during the stealth run
- October 23, 2026: MiniMax said open weights would arrive; license terms not disclosed at launch

## Access

API key at platform.minimax.io, the M Plan subscription, or the MiniMax Code desktop app. On OpenRouter as minimax/minimax-m3.1-flash.

## Related

- DeepSeek V4.1 Flash-1017: https://lab.awaara.cn/deepseek-v4-1-flash-1017/
- GPT-6 Luna pricing: https://lab.awaara.cn/gpt-6-luna-pricing/
- Claude Haiku 5.5 pricing: https://lab.awaara.cn/claude-haiku-5-5/