Qwen3-30B-A3B-Instruct: New model on ShareAI

Qwen3-30B-A3B-Instruct is now on ShareAI

We’re excited to add qwen3-30B-A3B-Instruct to the ShareAI model list—an instruction-tuned, Mixture-of-Experts 30B model with a long context window and strong upgrades in instruction following, reasoning, coding, and tool use.

Jump in via the Model Marketplace or try it instantly in the Chat Playground.

What it is (at a glance)

• Family: Qwen3 (open-weight), Instruct (non-thinking) variant
• Architecture: Sparse MoE; ~30B total parameters with only a fraction activated per token for efficient throughput
• Context: Long-context operation suitable for multi-file and large-document workflows
• Languages: Broad multilingual coverage for prompts and responses
• Agentic use: Optimized for tool calling; pairs well with agent frameworks

Why it matters

MoE efficiency without giving up capacity. qwen3-30B-A3B-Instruct delivers high quality while keeping per-token compute in check—ideal for apps that need strong reasoning and coding quality with predictable latency and spend.

Long context, practical workflows. A very large context window unlocks long-document QA, large-bundle retrieval, and “multi-file chat” experiences with fewer chunking compromises.

Good local story. Community tooling and quantized builds make local or small-server deployment feasible for prototyping and edge-adjacent use cases.

Benchmark & capabilities (highlights)

The latest Qwen3 refresh emphasizes better instruction following, long-tail knowledge, and alignment—plus tangible gains across math, code, and tool use versus prior non-thinking releases.

• Instruction following & reasoning: Upgrades across general tasks (math/science/code, tool calls).
• Agentic workflows: Robust tool-use patterns; designed to orchestrate multi-step actions.
• Multilingual: Strong performance across 100+ languages for prompts and tasks.

What you can build

• Large-document RAG & review: Process long PDFs, technical specs, and multi-file corpora in a single conversation.
• Agentic assistants: Connect search, code execution, and internal APIs to automate multi-step workflows.
• Developer copilots: Reasoning + coding strength for IDE chat, unit-test suggestions, and refactoring hints.
• Multilingual helpdesks: Serve global users with consistent intent understanding.

Get started on ShareAI

• Browse Models
• Open Playground
• Read the Docs
• Generate API Key
• Sign in or Sign up
• See Releases
• Provider Guide

ShareAI in one line: One API. 150+ AI Models — multi-provider routing, pay-per-token, 70% to GPUs.

Available now on ShareAI

Run it. Test it. Ship it.