How does MiniMax Sparse Attention (MSA) work?

MSA replaces full attention with KV-block selection to cut per-token compute at long context, roughly 1/20 the cost of the previous generation at 1M tokens, with substantially faster prefill.

How much does MiniMax M3 cost to use via API?

Pricing runs at $0.60 per million input tokens and $2.40 per million output tokens for contexts up to 512K, with prompt caching reads at $0.12 per million. A 50% launch discount was active for the first seven days.

When will the model weights be publicly available?

Model weights and a technical report will be published on Hugging Face and GitHub within the next ten days of launch, MiniMax says.

Is MiniMax M3 truly open source?

Open weight means the trained model parameters are made available for download and local deployment. Open source, in the stricter sense, means the training data, training code, and license terms also permit unrestricted commercial use. MiniMax has used a modified-MIT license for prior models, which is closer to open weight than to fully open source.

AI News06/02/26June 2, 20268 min

Jim Clyde Monge•Author

MiniMax M3 Is the First Open-Weights Model to Combine Coding, 1M Context, and Native Multimodality

Q: What makes MiniMax M3 different from other open-weights models?

These three capabilities — frontier coding, 1M-token context, and native multimodality — are now table stakes for closed-source frontier models. M3 is currently the first and only open-weight model to bring all three together.

MiniMax released M3 on June 1, 2026, stacking frontier coding performance, a million-token context window, and native multimodality into a single open-weights model for the first time.

Last updated on August 16, 2026

Chinese AI company MiniMax released M3, a new open-weight model that combines strong coding performance, native multimodality, and a one-million-token…

AI Summary

MiniMax released M3 on June 1, 2026, marking the first open-weights model to simultaneously combine frontier-level coding performance, a one-million-token context window, and native multimodality. The Chinese AI company positions M3 for long-horizon agentic workflows, complex coding tasks, and long-form video understanding, capabilities that until now existed only in closed-source models like GPT-5 and Gemini. On SWE-Bench Pro, M3 scores 59.0%, edging out GPT-5.5 at 58.6% while costing roughly 12 times less per input token. The technical foundation is MiniMax Sparse Attention, a new architecture that reduces per-token compute to approximately one-twentieth of its predecessor at million-token contexts. The model was trained natively on multimodal data from the start, rather than bolting vision onto an existing text system, enabling tighter integration across text, image, and video inputs. With open weights and a competitive price point, M3 gives developers a credible alternative to proprietary APIs for sensitive, long-context agentic work. Independent benchmark verification and public weight release remain pending, making those two milestones the clearest signals of whether M3 delivers on its considerable claims.

Chinese AI company MiniMax released M3, a new open-weight model that combines strong coding performance, native multimodality, and a one-million-token context window. That combination sounds like a checklist until you realize no open-weights model has managed all three at once. These three capabilities are now table stakes for closed-source frontier models. M3 is currently the first and only open-weight model to bring all three together.

The timing is pointed. Chinese open-weights models keep nibbling away at the leads of their US closed-source counterparts. M3 is the latest entrant, and MiniMax claims it is the first open-weights model to simultaneously deliver frontier-level coding, million-token context, and native multimodal capabilities. Until now, that combination had been the exclusive preserve of closed-source giants like GPT-5 and Gemini.

As someone who covers this space daily, M3 is the kind of release that actually warrants attention. Not because of the benchmark sheet alone, but because of what it means for developers who need long-context agentic workflows without routing sensitive code through a closed API.

What Is MiniMax M3?

MiniMax M3 is a multimodal foundation model from MiniMax. It accepts text, image, and video inputs and produces text output, with a context window of up to 1 million tokens and a guaranteed minimum of 512K. MiniMax positions it for long-horizon agent tasks, long-range coding, and long-form video understanding.

Over the last eighteen months, MiniMax built a solid reputation in the open-weights field, first with the MiniMax-01 series and then with the M1 and M2 reasoning models. With M3 the company raises the stakes and aims squarely at the tier of Western flagship models.

Coding and Agentic Performance

SWE-Bench Pro is the toughest version of SWE-Bench, the test that asks a model to fix real bugs and issues pulled from open-source software projects, complete with test execution to verify that the solution actually works. A score of 59.0% means M3 correctly resolves nearly six problems out of ten under conditions that trip up even the most expensive proprietary models.

The full benchmark picture across the agentic frontier:

SWE-Bench Pro: 59.0%, behind Claude Opus 4.7 (64.3%) but ahead of GPT-5.5 (58.6%) and Gemini 3.1 Pro (54.2%)
Terminal Bench 2.1: 66.0%, trailing GPT-5.5's 78.2% and Gemini 3.1 Pro's 70.0%
BrowseComp: 83.5, ahead of Opus 4.7 at 79.3
SWE-fficiency: 34.8% | KernelBench Hard: 28.8% | MCP Atlas: 74.2%

To get closer to real developer workflows, MiniMax built a simulator framework that mimics typical behavior patterns. These include refining requirements, discussing solution approaches, reacting to intermediate results, and carrying tasks across multiple contexts. This exposes the model to multi-turn collaboration during training, not just single, clearly defined prompts.

Up next

Google Lighthouse 13.3 Now Audits Whether AI Agents Can Actually Use Your Site

Lighthouse 13.3.0, released May 7, 2026, moved the Agentic Browsing category from experimental into the default config. That's a meaningful shift. Before…