MiniMax M3.1 Flash
A lightweight, high-speed text model released by MiniMax, targeting everyday development scenarios. It is designed for low latency and high response speed, making it suitable for high-frequency tasks such as rapid code fixes and similar lightweight workloads.
Core Facts
A context window of 1 million tokens is roughly equivalent to 700,000 to 800,000 Chinese characters, enough to hold an entire long novel or a large code repository.
M3.1 Flash can be called via the AI SDK, Chat Completions, Responses, and the Anthropic Messages API, with the model name minimax/minimax-m3.1-flash.
Maximum output tokens: MiniMax M3.1 Flash is 512K, DeepSeek V4.1 Flash 1017 is 384K, and Claude Haiku 5.5 is 128K.
Timeline (last 90 days)
MiniMax M3.1 Flash asserted in 6 impossible-version runs that the agreement would pass, even though at least two countries explicitly opposed it.
Hunyuan 4 scored 31.4% on Frontier Code 2.0, almost on par with Fibonacci 5.1; on AA Omniscience V1.1 its hallucination rate was 89%, second only to MiniMax M3.1 Flash.
Claude Pass is a $9.99-per-month subscription service offering open-source coding models such as DeepSeek V4.1 Flash, Kimi K3.1 Pro, GLM 5.3, Qwen 3.8 Max and MiniMax M3.1 Flash, with a call rate limit 2-5 times the usual, usable independently via an OpenAI-compatible API.
The open weights of MiniMax M3.1 Flash are set to be released on the 23rd, and the promotional free period runs until October 25.
MiniMax M3.1 Flash was officially released, with output speed exceeding 200 tokens per second, nearly three times faster than the previous generation M3, and about 70% cheaper.
MiniMax M3.1 Flash has officially launched after ending its preview, and "SpaceMoney Alpha," which previously appeared anonymously on leaderboards, was its early version.
MiniMax M3.1 Flash scored 61.7 on the computer operation benchmark Oathworld.
For Agent applications that require long context, multi-turn tool calling, code understanding, or computer operation capabilities, M3.1 Flash is an option worth evaluating.
The inability to turn off thinking depth may affect legacy applications that rely on fixed output formats or low-latency responses, so regression testing before the shutdown is recommended.
The official claim is that the overall cost of Agent-type tasks dropped by 70%.
40 more timeline events
All Facts (20)
A context window of 1 million tokens is roughly equivalent to 700,000 to 800,000 Chinese characters, enough to hold an entire long novel or a large code repository.
75%VerifiedM3.1 Flash can be called via the AI SDK, Chat Completions, Responses, and the Anthropic Messages API, with the model name minimax/minimax-m3.1-flash.
75%VerifiedMaximum output tokens: MiniMax M3.1 Flash is 512K, DeepSeek V4.1 Flash 1017 is 384K, and Claude Haiku 5.5 is 128K.
70%VerifiedM3.1 Flash's output speed reaches 200+ tokens per second.
65%VerifiedMiniMax M3.1 Flash offers a 1M-token ultra-long context window.
65%VerifiedThe official name of MiniMax M3.1 Flash is `MiniMax-M3.1-Flash`, positioned as a frontier multimodal coding model with a 1 million context window and adjustable thinking depth.
65%VerifiedMiniMax M3.1 Flash Preview emphasizes speed and can be used for quickly fixing bugs in code development.
65%VerifiedA Prompt Caching cache hit read costs only 0.1 times the base input price, while the initial write cost is 2 times the regular input price.
65%UnverifiedMiniMax M3.1 Flash uses a MoE architecture with approximately 428B total parameters and approximately 23B activated per token.
60%UnverifiedMiniMax M3.1 Flash defaults to the max setting when the effort field is not passed in the request, and the model always performs thinking.
60%UnverifiedWhen the reasoning_effort field is not passed, MiniMax M3.1 Flash runs at the max thinking level, taking an average of 56.3 seconds for a small repair task, which may trigger client timeouts tuned for fast models.
60%UnverifiedUsing DeepSeek V4.1 Flash 1017 directly in tools such as Claude Code, Codex, and Cursor requires installing the latest version of Vercel CLI and running the npx vercel ai-gateway setup initialization command.
60%UnverifiedMiniMax M3.1 Flash asserted in 6 impossible-version runs that the agreement would pass, even though at least two countries explicitly opposed it.
50%UnverifiedHunyuan 4 scored 31.4% on Frontier Code 2.0, almost on par with Fibonacci 5.1; on AA Omniscience V1.1 its hallucination rate was 89%, second only to MiniMax M3.1 Flash.
50%UnverifiedClaude Pass is a $9.99-per-month subscription service offering open-source coding models such as DeepSeek V4.1 Flash, Kimi K3.1 Pro, GLM 5.3, Qwen 3.8 Max and MiniMax M3.1 Flash, with a call rate limit 2-5 times the usual, usable independently via an OpenAI-compatible API.
50%UnverifiedMiniMax M3.1 Flash has five adjustable thinking-depth levels, and this capability cannot be turned off.
50%UnverifiedMiniMax M3.1 Flash scored 61.7 on the computer operation benchmark Oathworld.
50%UnverifiedFor Agent applications that require long context, multi-turn tool calling, code understanding, or computer operation capabilities, M3.1 Flash is an option worth evaluating.
50%UnverifiedMiniMax M3.1 Flash scored 73.7% on the complex coding benchmark Deeps V1.1.
50%UnverifiedIn the officially published comparison table of 15 benchmarks, MiniMax M3.1 Flash ranks first in only 3 and places in the top four in all the rest.
50%