Chinese AI startup MiniMax released its highly anticipated M3 large language model on Sunday evening Eastern time, pairing frontier-tier coding and agentic performance with a 1-million-token context window and native multimodality. The model undercuts leading U.S. proprietary models on cost by a dramatic margin.
Pricing starts at $20 per month under subscription plans, with API access at a discounted $0.3 per million input tokens and $1.20 per million output tokens for the next week. At full price ($0.6/$2.40 per million tokens), the model remains at just 8-20% the cost of comparable models from OpenAI, Google, and Anthropic, according to the company.
MiniMax also announced plans to release the model under an open-source license with open weights within the next 10 days, allowing enterprises to download and customize it free of charge. This strategy mirrors a broader trend among Chinese AI firms competing on openness and cost.
The M3's performance claims are notable: MiniMax says it eclipses GPT-5.5 and Gemini 3.1 Pro on key benchmarks, though independent verification is still pending. If validated, the model could challenge the dominance of U.S. frontier labs, particularly in cost-sensitive enterprise deployments.
Founder background was not detailed in the source, but the release signals a significant escalation in the cross-border AI race. Counter_argument: Independent benchmarks have not yet confirmed MiniMax's performance claims, and the model's open-source release is still days away, leaving room for potential discrepancies in real-world testing.
a_context: This brief is based solely on VentureBeat's reporting. Specific performance claims are attributed to MiniMax and have not been independently verified. Pricing and benchmark comparisons are taken directly from the source.