MiniMax slips 1M-context model into Code

MiniMax has put M3.1-Flash-Preview into MiniMax Code with a 1-million-token context window and five reasoning-effort levels, but no public price or model card. The MiniMax Code report says the release is in-product only, with no public API endpoint or benchmark report yet. Treat it as a bounded coding-agent experiment until those missing artifacts arrive; teams comparing workhorse models can keep it beside Google’s fast-moving Flash price war without binding production traffic to an undocumented preview.