MiniMax M3.1 Flash vs. DeepSeek V4.1 Flash in a Hands-On Agent Test
大肥鱼遇劲敌：MiniMax M3.1 Flash 拿下 Agent 任务第一，我们实测了
- Who
- MiniMax
- What
- published benchmark results M3.1 Flash results on Automation-Bench, BankerToolBench, and other benchmarks
- Quote
在此次发布的 Automation-Bench v1.0.6 模型对比中，M3.1 Flash 超过 DeepSeek，位列第一；BankerToolBench 则拿到了 76.2%，与 Opus 5 并列第一。
What happened
Leiphone tested MiniMax M3.1 Flash and DeepSeek V4.1 Flash on building an interactive 3D campus from a hand-drawn plan and building dimensions. Both setups met the main requirements, but MiniMax Code plus M3.1 Flash took 55.4 minutes versus about 121 minutes for OpenCode plus DeepSeek V4.1 Flash, with model charges around 6.5 yuan each under the pricing used. Published benchmarks placed M3.1 Flash ahead on complex agent tasks and DeepSeek ahead on software engineering.
Why it matters
The campus test offers a concrete comparison of elapsed time and model charges across two coding setups, illustrating why agent workflows benefit from measuring whole-task costs.
The story so far
MiniMax M3.1 Flash ranked first on Automation-Bench v1.0.6 in results published by MiniMax, while a hands-on comparison by Leiphone found that a setup using the model completed an interactive 3D campus in less than half the time of a setup using DeepSeek V4.1 Flash. Both setups met the main requirements, which involved building the campus from a hand-drawn plan and building dimensions. MiniMax Code with M3.1 Flash took 55.4 minutes, compared with about 121 minutes for OpenCode with DeepSeek V4.1 Flash. Model charges were around 6.5 yuan for each setup under the pricing used. The comparison involved different coding tools as well as different models. The published benchmarks placed M3.1 Flash ahead on complex agent tasks and DeepSeek ahead on software engineering; MiniMax also reported a 76.2% score for M3.1 Flash on BankerToolBench, tying Opus 5 for first place.
Keep reading
Latest daily · Issue 28Mistral AI releases Voxtral 3 speech models and a voice agent SDKTrending now
- 1Mistral Releases Voxtral 3 Speech Models, Voice Agents API and LiveKit-Based SDK
- 2DeepSeek Open-Sources DeepInfer and API Components, With Model and Hardware Limits
- 3OpenAI Acquires Orchestra Bio to Expand AI Tools for Drug Development
- 4Google Rolls Out Gemini Personal Intelligence to Android Auto, With Google Built-In Cars to Follow
Tomorrow at 08:00, a briefing like this arrives in your inbox.
The same kind of briefing as the one above: a few stories worth reading. Free. Mail starts after you confirm, and every letter has a one-click unsubscribe. Read a past issue · 2026-10-21