Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
OpenCode Zen logo

Model details

MiniMax-M3 Free

This free-tier route was made broadly available during a limited promotion, offered at no charge across input, output, and cache reads by gating the underlying frontier model behind gateways such as GMI Cloud and pricing directories like Command Code. The same intelligence index and coding score attach to the free alias as to its paid sibling, suggesting that the promotion is purely a billing and routing convenience layered over the base architecture rather than a trimmed-down variant. Practical use fits coding harnesses, agent loops, and multimodal exploration where cost matters more than throughput, since the free alias produces identical reasoning, tool calling, and native multimodality behavior to the underlying model.

The model is positioned as a frontier-tier coding and agentic system with native multimodality, and its open-weight status makes it attractive for self-hosting once the free window closes. Community tooling has caught up quickly, with coding agents surfacing it as a selectable option alongside new reasoning effort controls, indicating that developers can now tune depth-of-thought alongside the model's existing tool-calling strengths. On OpenCode Zen itself the alias is already flagged as deprecated, so it serves best as a low-risk way to evaluate capabilities and benchmark workflows before transitioning to a paid route or self-hosted deployment for ongoing production use.

OpenCode Zenminimax-m3-freeminimax-m3-freedeprecated

Quick Info

Powered by
Provider
OpenCode Zen
Model key
minimax-m3-free
Release date
May 31, 2026
Last updated
May 31, 2026
Knowledge cutoff
2025-01
AI SDK package
@ai-sdk/anthropic
Input modalities
Output modalities
Capabilities

Cost

A provider subscription or plan supersedes token-based pricing for this model.

Limits

Output tokens
32,000 tokens
Context window
200,000 tokens

Latest news about MiniMax-M3 Free

OpenCode Zen

Coverage

A June 30, 2026 Medium explainer by developer Rowan Blackwoon reports that MiniMax M3 was released on June 1, 2026 as an open-weight model combining frontier-level coding, a context window of up to 1,000,000 tokens, and native multimodality for image and video input. The post highlights that the same release can also o The same piece walks through MiniMax's reported benchmarks and the architectural motivation behind M3's long-context efficiency, positioning the release as a notable step in the open-weight race following releases like Qwen 3.7. It discusses practical implications for builders wiring M3 into applications, including API

OpenCode Zen

CoverageBenchmark

According to its fact-checked reference page (last checked August 26, 2026), MiniMax M3 is described as an open-weight, native multimodal Mixture-of-Experts model released on June 1, 2026, with approximately 428 billion total parameters and about 23 billion active parameters per token. It accepts text, image, and video The same page details that the model uses MiniMax Sparse Attention and supports a self-hosting workflow, with weights distributed on Hugging Face and deployment code in official GitHub repositories under the MiniMax Community License. It also lists standard API pricing of $0.30 per million input tokens and $1.20 per mi

OpenCode Zen

CoverageBenchmark

A June 28, 2026 review article reports that MiniMax M3 launched on June 1, 2026 as the first open-weights model to combine frontier-level coding, a one-million-token context window, and native text/image/video understanding in a single freely downloadable release. It cites MiniMax vendor-reported benchmarks placing M3 The same piece describes MiniMax (officially Xiyu Technology) as a Shanghai-based AI lab founded in 2021 and listed on the Hong Kong Stock Exchange since January 2026, framing M3 as the lab's flagship model. It reports approximately 428 billion total parameters for the downloadable weights and walks readers through acc

OpenCode Zen

CoverageBenchmark

The OpenRouter listing explicitly names the "MiniMax M3 (free)" variant with model id minimax/minimax-m3:free, confirming the existence of a free tier of the MiniMax M3 model family. Per the supplied page excerpt, MiniMax-M3 is described as a multimodal foundation model from MiniMax that accepts text, image, and video The same listing reports benchmark figures for the free variant, including GPQA Diamond scores from multiple providers (Together 89.1%, ModelRun 88.6%, Minimax 90.9%, DeepInfra 89.2%, SambaNova 80.6%) and TAU-Bench scores (Together 76.5%, ModelRun 74.1%, Minimax 70.9%, DeepInfra 71.1%), indicating consistent capability

OpenCode Zen

Coverage

The dedicated model information page for MiniMax-M3 Free on OpenCode Zen confirms the model key "minimax-m3-free" with a release and last-updated date of May 31, 2026, and a knowledge cutoff of 2025-01. The page lists input modalities of text, image, and video with text output, and the AI SDK package is shown as @ai-sd Capabilities surfaced on the page include reasoning, tool calling, open weights, and temperature controls, with a 200,000-token context window and a 32,000-token output limit. Pricing is described as provider-subscription-based rather than token-based, and the page itself flags the model as "deprecated" on the provider

OpenCode Zen

CoverageRelease Notes

Command Code's v0.1.19 Desktop changelog (Aug 25, 2026) announces that MiniMax M3 and MiniMax M2.7 are being made free through September 5, 2026 within that third-party coding CLI/Desktop product, with a clearly displayed FREE badge added to free models in the model picker. This is direct, timely third-party evidence t The same v0.1.19 entry also ships a behavioral fix — "MiniMax M3 now handles task lists reliably" — alongside other improvements such as adding GLM-5.3 Flash with image support and a 1M-token context window, and various Windows installer and session-handling fixes. The free window is explicitly time-bounded (through Se

OpenCode Zen

CoverageBenchmark

Command Code's changelog entry dated August 25, 2026 (CLI v1.33.0) explicitly lists "Add MiniMax M3 Free and M2.7 Free" as a new feature, providing direct, dated technical confirmation that MiniMax M3 Free is being surfaced as a named, distributable model across third-party AI coding harnesses. This is the strongest in The entry appears alongside other model additions such as Ox Alpha and DeepSeek V4 Flash Vision, reinforcing that MiniMax M3 Free is being treated as a standard model option in agentic CLI workflows. However, the Command Code changelog does not specify routing, rate limits, or whether the integration pulls from opencod

OpenCode Zen

CoverageRelease Notes

Zed's preview releases page documents a recent improvement to OpenCode support that directly affects the MiniMax-M3 Free model served through OpenCode Zen. The changelog notes "updated OpenCode Zen models (added Fable 5, Sonnet 5, GLM 5.2, Kimi K2.7 Code, and Minimax M3), added reasoning effort levels," confirming that The same Zed release entry also notes that LLM providers, external agents, and MCP servers were moved into the settings editor, giving developers a centralized place to manage OpenCode Zen and its Minimax M3 model alongside other providers. Combined with the new reasoning effort control, the change means Zed users can

Videos about MiniMax-M3 Free