Qwen3 Coder Next
Qwen3 Coder Next is the forward-looking iteration of Alibaba's agentic coding model line, designed for developers who want rolling updates to code generation and software engineering capabilities as Alibaba publishes new Qwen3-Coder checkpoints.
import { streamText } from 'ai'
const result = streamText({ model: 'alibaba/qwen3-coder-next', prompt: 'Why is the sky blue?'})Playground
Try out Qwen3 Coder Next by Alibaba. Usage is billed to your team at API rates. Free users (those who haven't made a payment) get $5 of credits every 30 days.
Providers
Route requests across multiple providers. Copy a provider slug to set your preference. Visit the docs for more info. Using a provider means you agree to their terms, listed under Legal.
P50 throughput on live AI Gateway traffic, in tokens per second (TPS). Visit the docs for more info.
P50 time to first token (TTFT) on live AI Gateway traffic, in milliseconds. View the docs for more info.
Direct request success rate on AI Gateway and per-provider. Visit the docs for more info.
More models by Alibaba
| Model |
|---|
| 240K | 3.5s | 45tps | $1.30/M | $7.80/M | Read: $0.26/M Write: $1.63/M | — | — |  | 04/20/2026 | ||||
| 1M | 0.2s | 99tps | $0.50/M | $3.00/M | Read: $0.1/M Write: $0.63/M | — | — |  | 04/02/2026 | ||||
| 1M | 1.8s | 82tps | $0.10/M | $0.40/M | Read:$0.0/M Write:$0.13/M | — | — | | 02/24/2026 | ||||
| 1M | 1.3s | 110tps | $0.40/M | $2.40/M | Read: $0.04/M Write: $0.5/M | — | — | | 02/16/2026 | ||||
| 262K | 0.1s | 1039tps | $0.07/M | $0.46/M | Read:$0.6/M Write:— | — | — |    +1 | 04/01/2025 | ||||
| 131K | 0.1s | 359tps | $0.10/M | $0.30/M | Read:$0.14/M Write:— | — | — |  +1 | 04/01/2025 |
About Qwen3 Coder Next
Qwen3 Coder Next is the forward-tracking identifier in the Qwen3-Coder model series: it resolves to whichever checkpoint Alibaba assigns to the rolling Qwen3-Coder line. Rather than referencing a fixed model version, it tracks forward progress as the team continues refining the model's coding, reasoning, and agentic capabilities.
The Qwen3-Coder family was built on the recognition that production software engineering requires more than code completion. It requires models that can plan across multi-step tasks, use tools, interpret execution results, and maintain coherence across extended sessions. The "Next" designation points at the rolling checkpoint for that line.
For developers building applications that need to follow ongoing improvements in open agentic coding capability, automated software development tools, AI pair programmers, or repository-level automation, Qwen3 Coder Next is the identifier that tracks those improvements without requiring code changes when underlying model weights are updated. The tradeoff is that behavior may shift between underlying versions; applications that need strict reproducibility should pin to a versioned model identifier instead.
The core capabilities of the Qwen3-Coder line (deep code understanding, multi-turn agentic task execution, tool calling, and long context support) carry through to this variant.
What To Consider When Choosing a Provider
- Configuration: Because this model tracks a moving development line, provider availability may shift as new checkpoints are deployed. AI Gateway's failover mechanism helps maintain continuity during transitions.
- Zero Data Retention: AI Gateway supports Zero Data Retention for this model via direct gateway requests (BYOK is not included). To configure this, check the documentation.
- Authentication: AI Gateway authenticates requests using an API key or OIDC token. You do not need to manage provider credentials directly.
When to Use Qwen3 Coder Next
Best For
- Applications where recency of capability matters: For teams who want their coding AI to automatically pick up new Qwen3-Coder checkpoints as Alibaba releases them, this identifier avoids manual version updates
- Exploratory and research-oriented coding tools: When you're building and iterating rapidly and want whatever checkpoint this identifier resolves to at each stage of development rather than maintaining a pinned version
- Agentic software engineering workflows: The Qwen3-Coder foundation (long-horizon agent RL, tool calling, plan-execute-debug loops) applies to this variant, making it suitable for multi-step coding automation
- Prototyping new coding AI capabilities: Testing what each new checkpoint can do before deciding whether to pin a specific version for production is a natural use of a "next" model identifier
Consider Alternatives When
- Reproducibility is required: "Next" model identifiers may point to different underlying checkpoints over time. For production systems where consistent behavior is a compliance or quality requirement, use a specific versioned model ID
- You need guaranteed stability across deployments: Applications in production where behavior consistency is critical (code review systems, automated test generation with known expectations) should pin to a fixed model version
- Cost predictability is important: As model versions update, performance characteristics including speed and token usage may change. Fixed model versions are easier to budget around
Conclusion
Qwen3 Coder Next gives development teams a stable API integration that tracks ongoing work in Alibaba's agentic coding line. The practical advantage is receiving checkpoint updates without migration effort; the practical caution is that behavior isn't frozen. For most exploratory and rapid-iteration use cases, this tradeoff is favorable. AI Gateway adds reliability across togetherai, bedrock on top of that access.
Frequently Asked Questions
What does "Next" indicate in the model name?
The "Next" suffix designates this as a forward-tracking model identifier. It resolves to Alibaba's rolling Qwen3-Coder checkpoint rather than a single fixed snapshot. As Alibaba releases improved versions, this identifier updates to point at them.
Will my application behave differently after an underlying model update?
Potentially yes. "Next" identifiers trade reproducibility for recency. If consistent behavior is required, for testing, compliance, or quality assurance, pin your application to a specific versioned model identifier instead.
What are the core coding capabilities shared with the rest of the Qwen3-Coder family?
Core capabilities include multi-language code generation, multi-step agentic task execution, tool calling, and extended context handling. See https://www.together.ai/models/qwen3-coder-next for the full capability overview.
How large is the context window for Qwen3 Coder Next?
The context window is 256K tokens. The Qwen3-Coder architecture natively supports long contexts, with YaRN extrapolation available to extend beyond native limits.
Is there a latency or throughput difference between this and pinned Qwen3-Coder versions?
This depends on which specific checkpoint "Next" points to at any given time. Live metrics are shown on this page and reflect real-time traffic conditions.
Can Qwen3 Coder Next be used for browser-use and tool-use agentic benchmarks?
The Qwen3-Coder family achieved strong reported results on open models for Agentic Coding, Agentic Browser-Use, and Agentic Tool-Use. This variant, tracking the rolling Qwen3-Coder line, should carry those agentic characteristics forward.
What should I do if I need reproducible outputs from this model?
Switch to a versioned model ID for that workload. You can use AI Gateway to access both: the "Next" identifier for ongoing development and exploration, and a pinned version for the production component that requires stability.