MiniMax M3.1 Flash cuts AI coding costs below $0.30 a task

What happened

MiniMax has released M3.1 Flash, which scores 42 on the Artificial Analysis Intelligence Index. That is 13 points above its predecessor, and according to The Decoder, it does so at half the cost.

On coding, the model reaches 73.7 percent on the DeepSWE v1.1 benchmark for under $0.30 per task. The Decoder reports that this matches Deepseek V4.1 Flash 1017 on both performance and price, placing the launch within a broader pricing race among Chinese AI developers.

Why it matters

For organisations buying AI capability, the headline is less about a leaderboard position and more about cost per completed task. When comparable coding performance is available for a few tenths of a dollar per task, the economics of automating routine development, support and analysis work shift. Decisions that once needed careful budgeting can start to look like operational defaults, which changes how teams plan pilots and how they judge return.

The pattern also matters for leaders responsible for service and digital operations. As several providers converge on similar performance and price, a model stops being the differentiator on its own. Advantage moves to how well capability is integrated into workflows, how reliably it performs in production, and how thoughtfully it is designed into the experiences customers and employees actually use.

By the numbers

• 42 on the Artificial Analysis Intelligence Index, 13 points above the M3.1 Flash predecessor
• 73.7 percent on the DeepSWE v1.1 coding benchmark
• Under $0.30 per task on the DeepSWE v1.1 coding benchmark

The Renascence take

Most coverage of this launch will read as a price story. The more useful reading is about what happens to customer experience when a capability becomes cheap enough to deploy almost anywhere.

Most readers will fixate on the sub-dollar price tag and miss that falling unit costs turn AI from a strategic purchase into a commodity input. The behavioural principle underneath is that when a capability's price collapses, value shifts to the moments where it removes friction for a real person, and those moments are rarely obvious from a benchmark. A customer-obsessed operator should stop asking which model is cheapest and start mapping the journeys where a near-zero-cost task would genuinely reduce waiting, effort or errors. Then test those changes with real customers and employees before committing budget, because the experience gain is what justifies the spend.