| Model profile | ||
|---|---|---|
| Model | Size | Approximate size |
| MiniMax M3Model source: provider1,048,576 tokens max native context | 428Bactive 23B | 256.8 GB4-bit quantization |
| Compatible hardware Estimated speeds, not benchmark results: calculated from memory bandwidth and model size. Real results can differ significantly because there is no precise formula for deriving LLM generation speed from hardware specifications alone. | ||||
|---|---|---|---|---|
| Hardware | Memory | Bandwidth | vLLM | Estimated generation |
| DGX H200 | 1128 GBGPU HBM3e | 4800 GB/s | yes | 119 tok/s4-bit quantization |
| DGX Station | 748 GBCoherent Memory | 7100 GB/s | yes | 163 tok/s4-bit quantization |
| ET900N G3 | 748 GBCoherent Memory | 7100 GB/s | yes | 163 tok/s4-bit quantization |
| DGX B200 | 1440 GBGPU HBM3e | 8000 GB/s | yes | 178 tok/s4-bit quantization |
| GB200 NVL72 | 13400 GBGPU HBM3e | 8000 GB/s | yes | 178 tok/s4-bit quantization |
| GB300 NVL72 | 20000 GBGPU HBM3e | 8000 GB/s | yes | 178 tok/s4-bit quantization |
GLM 5.2, Qwen 3.7, and MiniMax m3 are compared on real coding tasks performed live within the Hermes Agent environment.
Tests MiniMax M3 in Claude Code UltraCode Mode against GPT-5.5 via large codebase review, Pi /goal extension build-off, and Splunk cyber CTF challenge.
Tests MiniMax M3 within MiniMax Code, demonstrating its role as an agentic coding workflow with features like multi-agent teams, output verification, and automated task execution.
Evaluates MiniMax M3 through browser workflows, C++ game development, scene generation, multimodal coding, simulators, frontend design, and drum kit generation.
Coding challenges test MiniMax-M3 versus Qwen3.6 27B using an iPhone-style UI, ragdoll physics simulator, and a fake Qwen3.7 30B release page to evaluate usability and polish.
Compares MiniMax M3 pricing against Claude Sonnet 4.6, Claude Opus 4.8, and GPT-5.5 for AI coding and testing workflows.
Reviews MiniMax M3, demonstrating its frontier-level performance on specialized tasks such as coding and agentic work.
Tests MiniMax M3 on SWE-Bench Pro, BrowseComp, SVG-Bench, KernelBench Hard, OSWorld Verified, frontend tasks, Three.js, SVG animation, and agentic workflows against Opus 4.7 and GPT-5.5.