The announcement reports 73.7% on DeepSWE v1.1 and 61.7 on OSWorld V2. It lists prices of $0.15 per million input tokens, $0.60 per million output tokens, and $0.015 per million cached input tokens, and says agent-task usage costs 70% less than its predecessor. These are company-published claims, not independently verified results.
Implication: Builders can compare the published rates and benchmark claims with the cost and completion quality of their own coding and computer-use tasks.
Thesis
A low-cost model release matters when teams can test the price-performance tradeoff in their own agent workflows.
Observed facts
Mechanism
M3.1 Flash pairs reported benchmark performance with published per-token prices and access through three MiniMax channels, making an initial evaluation possible without waiting for the announced later release of model weights.
Why now
MiniMax says the model is available now through its API, M Plan, and MiniMax Code. That enables teams to test the stated economics directly, while the launch is current.
Implications
Builders can compare the published rates and benchmark claims with the cost and completion quality of their own coding and computer-use tasks.
Evidence boundary
The supplied source is MiniMax’s announcement; it gives company-reported benchmark scores and pricing but no independent replication or detailed evaluation methodology in the supplied material.
Watch conditions
Independent evaluations and real workflow tests should show whether the reported capability and cost translate into reliable completed tasks.