MiniMax releases M3.1 Flash, formerly Space Bunny
MiniMax releases M3.1 Flash, formerly Space Bunny
MiniMax has officially introduced M3.1 Flash, previously known as Space Bunny. The model is reported to generate more than 200 tokens per second and outperform other Chinese models in benchmarks.
MiniMax has officially introduced its new model, M3.1 Flash. The release is the public version of what was previously known as Space Bunny, an early iteration of the model.
According to the available characteristics, M3.1 Flash looks promising:
- It can generate more than 200 tokens per second.
- It reportedly outperforms other Chinese models in benchmark tests.
The model appears to focus on speed and competitive performance, which may be important for applications where latency and throughput matter.
Why it matters
AnalysisFast generation and strong benchmark results could make M3.1 Flash useful for latency-sensitive AI applications. The release also shows continued competition among large language models outside the US.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Claude Sonnet 4.6 vs GPT-5.5 vs Gemini 3.1 Pro: 2026 model comparison
Every few weeks, a new model comes out, and every time the same questions arise: should I switch? what changed what to pay for?
Anthropic Reportedly Preparing Fable 5.5 Model
Insiders who tested a new Anthropic model say it better understands user intent from short prompts and is especially strong at finding bugs and vulnerabilities.
DeepSeek releases DeepInfer, an open-source inference framework
DeepSeek has published DeepInfer, an Apache-2.0 inference framework used in its production services, with optimized kernels, MoE parallelism, speculative decoding, deterministic mode, and structured outputs.
Part I. AI Agent Architecture: The Foundations
The eleven foundational building blocks of modern autonomous agents: model taxonomy, the agentic runtime loop, advanced prompting, distillation, skills architecture, memory systems, function calling, and FinOps.
Agent Benchmarks: GAIA and SWE-bench
A new generation of rigorous practical benchmarks for evaluating the autonomy of artificial intelligence. GAIA tests multimodal tasks in the open web with files and a browser, while SWE-bench assesses agents' ability to find and fix real bugs in large GitHub repositories.
Agent Evals & SWE-bench Benchmarking
A methodology and infrastructure for systematically measuring the reliability, accuracy, and safety of AI agents through synthetic tests, SWE-bench, and headless repository simulations.