⚡️ Step 3.5 Flash is coming: Fast Enough to Think. Reliable Enough to Act！
We’re dropping our most capable open-source foundation model yet. Frontier reasoning meets extreme efficiency.
It leverages a sparse Mixture of Experts (MoE) architecture, 196B total → 11B active.
Key Capabilities:
✅Reasoning at Speed: MTP-3 powered throughput at 100–300 tok/s (350 tok/s peak for single-stream coding tasks).
✅Agentic Power: ⚡️ 74.4% SWE-bench Verified ⚡️ 51.0% Terminal-Bench 2.0. Proven stability for complex, long-horizon tasks.
✅256K Efficient Context: 3:1 SWA ratio + Full Attention. Massive datasets or long codebases support with minimal overhead. Consistent performance, hybrid efficiency.
✅Local-First Deployment: Optimized for Mac Studio M4 Max, NVIDIA DGX Spark. Secure, private, and frontier-capable. Your data, your hardware, your agent.
You can try Step 3.5 Flash right now:
👉 OpenRouter: https://t.co/MRDG3R4F64
👉 GitHub: https://t.co/qqmpslxK9t
👉 HuggingFace：https://t.co/y70Qzsc8DE
👉 Blog：https://t.co/rMH67kYmvA
👉 ModelScope: https://t.co/T4W7we5GEr
🌌 The Next：Step 4 training is officially LIVE！
We're calling on the world's boldest builders to co-creat the Step 4 right now. Let's define the Agentic Era together!
Join our Discord：
Try Step 3.5 Flash on NV NIM👉
https://t.co/2F1cI15Zl3
mixture-of-experts — A model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
SWE-bench — The standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
Why it matters
A 196B mixture-of-expertsA model built from many specialist sub-networks where only a few activate per token, giving big-model capability at small-model running cost.Full definition → activating 11B parameters posts 74.4% SWE-benchThe standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.Full definition → and 51.0% Terminal-Bench 2.0 at 100 to 300 tokens per second, with open weightsA model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.Full definition → and local deployment on Mac Studio M4 Max and DGX Spark.
article👏🏻Congratulations！Step3-VL-10B was selected for HuggingFace Daily Papers…
postStep-Audio-R1.1 opens weights for a speech model that reasons in real time
