M

MiniMax M3.1 Flash

minimax · Veröffentlicht Okt. 2026

Bedingt

MiniMax M3.1 Flash is a 1M-context multimodal coding and agent model at $0.15 input and $0.60 output per 1M tokens, and MiniMax says its agentic usage cost is 70% lower than its predecessor's. The deciding caveat is evidence: the launch post's scores are MiniMax's own, it names no comparison model, and the weights are promised only for later. Worth a trial for cost-sensitive agent and computer-use workloads, not yet for anyone who needs self-hosting or independent scores.

Ist es das Richtige für dich?

Gut für

  • Agent and computer-use work on a budget: $0.15 / $0.60 per 1M tokens with cached input at $0.015
  • Long-context coding and tool use: 1M context window, tunable thinking depth and interleaved thinking with tool calls
  • Fast agent loops: MiniMax reports output speeds of 200+ tokens per second and a median above 170

Nicht geeignet für

  • Self-hosting today: the post says 'Model weights will be released later', with no date
  • Decisions that need independent benchmark evidence: the scores in the launch post are MiniMax's own, and the pages we read name no third-party run

Leistung nach Aufgabe

Agentic tool use

Good

Documented as an agentic model with interleaved thinking and tool calling over Anthropic and OpenAI SDKs; MiniMax reports it ranks among the top performers on AutomationBench-AA (vendor claim)

Computer use

Good

MiniMax reports 61.7 on OSWorld V2 and 29.6 on Agents' Last Exam, first among Chinese models on both (vendor claim)

Software engineering

Good

MiniMax reports 73.7% on DeepSWE v1.1 (vendor claim, comparison models unnamed)

Preise

Eingabe

$0.15 / 1M tokens

Ausgabe

$0.60 / 1M tokens

Kontext

1M context; cached input $0.015 / 1M

Alle Preise ansehen

Benchmarks

BenchmarkWertQuelle
DeepSWE v1.173.7% Quelle
OSWorld V261.7 Quelle
Agents' Last Exam29.6 Quelle
BankerToolBench76.2 Quelle
CyberGym82% Quelle

Noch keine Urteilsänderungen

Die Uhr läuft ab dem ersten Tag. Änderungen erscheinen hier, sobald sich unser Urteil weiterentwickelt.

Prüfprotokoll

Noch keine Prüfungen

Wir haben für diesen Eintrag noch keine Prüfung erfasst. Sobald eine Prüfung läuft, erscheint ihr Verlauf hier.

Wie wir bewerten