MiniMax M3
minimax-m3MiniMax M3 is MiniMax's open-weight, native multimodal MoE vision-language model (~428B total, ~23B active parameters) with a 1M-token context window, powered by the MSA architecture for frontier-level reasoning, long-horizon coding, and agentic tool use across text, image, and video input.
Best for
Characteristics
Modalities
Use cases
Anthropic
Yes
Chat Completions
Responses Api
Yes
Provider
MiniMax
Context Window
1M
Available Access
Prices shown per 1M tokens.
Input
Rp. 4.310
$0.24 USD
Output
Rp. 17.240
$0.97 USD
Cache
Rp. 1.862
$0.1 USD
All prices are estimates and subject to change.
Similar Models
Related models selected by the catalog administrator.
minimax-m3-freeMiniMax M3 Free is the no-cost access tier to MiniMax's native multimodal MoE model, providing the same 1M-context reasoning, coding, and agentic capabilities as the paid M3 at no charge.
qwen3.8-flashQwen3.8 Flash is the official fast-tier API version of Alibaba's Qwen3.8 generation, based on the Qwen3.8-Flash-Next architecture (an experimental Qwen4-preview design) and offering low-latency, high-throughput multimodal reasoning with 1M-token context and built-in tools.
deepseek-v4-proDeepSeek's V4 Pro model, the stronger, reasoning-optimized flagship tier of the DeepSeek-V4 generation aimed at complex problem-solving, coding and agentic tasks, with a 1M-token context.