← All models

MiniMax M3.1 Flash

Language Available Full comparison ↗

minimax/m3.1-flash · by MiniMax · decoder-only-transformer

Pricing — 1 offering(s)

MiniMax

About this offering: MiniMax M3.1 Flash (API identifier: "MiniMax-M3.1-Flash"), MiniMax's low-cost Flash model for agentic work, coding and computer use, released 2026-10-19. 1,000,000-token context window; text, image and video input. Flat pricing at every context length ($0.15 input / $0.60 output / $0.015 cached input per 1M tokens), half M3's ≤512k rate. Thinking depth is tunable with effort (low to max, default max). Model weights aren't released yet; MiniMax says they'll follow later.

Input tokens

  • $0.15 / 1M tokens (input) Current 2026-10-20 → present

Cached input tokens

  • $0.015 / 1M tokens (cached input) Current 2026-10-20 → present

Output tokens

  • $0.60 / 1M tokens (output) Current 2026-10-20 → present
    Thinking is on by default (effort defaults to max), and thinking tokens are output tokens, so lower effort…

    Thinking is on by default (effort defaults to max), and thinking tokens are output tokens, so lower effort settings cut output cost.

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

See how MiniMax M3.1 Flash fits into a cost-aware routing setup

See how →

Capability profile

How MiniMax M3.1 Flash rates across core capability dimensions, with the task-level evidence behind each rating.

reasoning Strong

Thinks before answering by default, with tunable effort from low to max (default max). Vendor positioning; no independent reasoning benchmark recorded yet.

coding Strong

DeepSWE v1.1 (complex software tasks validated by tests): 73.7%. Vendor-reported (MiniMax launch post); not independently confirmed.

tool use Strong

MiniMax docs: "an Agentic Model with exceptional Tool Use capabilities", with interleaved thinking. Vendor-reported OSWorld V2 61.7 and BankerToolBench 76.2 (launch post).

context window Strong

1,000,000-token context window, with flat pricing at every context length.

speed Strong

MiniMax reports 200+ tokens per second output, with a median (P50) of 170+, and "nearly three times faster" online inference than its predecessor. Vendor-reported.

cost efficiency Strong

$0.15/$0.60 per 1M tokens with cached input at $0.015: half MiniMax M3's ≤512k rate and the same as DeepSeek V4.1-Flash's off-peak price.

Ratings are estimated — limited independent data is available for this model.

Operator guidance

Route here for agentic, coding and computer-use work at Flash-tier cost: $0.15/$0.60 per 1M tokens, flat at every context length, with a 1M-token window. Thinking defaults to max effort and thinking tokens bill as output, so set a lower effort for simple tasks. MiniMax M3 remains the flagship at twice the price and is the only one with a Priority service tier. Capability claims are MiniMax's own so far, so check results on your task.

Use cases

Limitations

Citations