Qwen: Qwen3 Coder Next

Text input Text output
Author's Description

Qwen3-Coder-Next is an open-weight causal language model optimized for coding agents and local development workflows. It uses a sparse MoE design with 80B total parameters and only 3B activated per token, delivering performance comparable to models with 10 to 20x higher active compute, which makes it well suited for cost-sensitive, always-on agent deployment. The model is trained with a strong agentic focus and performs reliably on long-horizon coding tasks, complex tool usage, and recovery from execution failures. With a native 256k context window, it integrates cleanly into real-world CLI and IDE environments and adapts well to common agent scaffolds used by modern coding tools. The model operates exclusively in non-thinking mode and does not emit <think> blocks, simplifying integration for production coding agents.

Key Specifications
Cost
$$$$
Context
262K
Parameters
80B (Rumoured)
Released
Feb 03, 2026
Speed
Ability
Reliability
Supported Parameters

This model supports the following parameters:

Response Format Presence Penalty Tools Top P Frequency Penalty Max Tokens Structured Outputs Seed Tool Choice Stop Temperature
Features

This model supports the following features:

Tools Structured Outputs Response Format
Performance Summary

Qwen3-Coder-Next, released by qwen on February 3, 2026, is an open-weight causal language model designed for coding agents and local development. With a sparse MoE design (80B total parameters, 3B activated per token), it achieves performance comparable to models with significantly higher active compute, making it suitable for cost-sensitive deployments. The model exhibits a native 256k context window and operates exclusively in non-thinking mode, simplifying integration. In terms of overall performance, Qwen3-Coder-Next tends to have longer response times, ranking in the 19th percentile across five benchmarks. However, it offers competitive pricing, placing in the 56th percentile. A significant strength is its exceptional reliability, demonstrating a 100% success rate across all benchmarks, indicating consistent and usable responses. Benchmark results highlight several key areas. The model shows strong performance in Ethics, achieving perfect accuracy and being noted as the most accurate model at its price point and among models of similar speed. It also performs well in General Knowledge (99.0% accuracy) and Hallucinations (94.0% accuracy), indicating a good grasp of factual information and an ability to acknowledge uncertainty. While its Instruction Following accuracy is moderate at 59.6%, its Email Classification accuracy is solid at 97.0%. The primary weakness lies in its speed, consistently ranking low in duration across most benchmarks.

Model Pricing

Current Pricing

Feature Price (per 1M tokens)
Prompt $0.2
Completion $1.5

Price History

Available Endpoints
Provider Endpoint Name Context Length Pricing (Input) Pricing (Output)
Novita
Novita | qwen/qwen3-coder-next-2025-02-03 262K $0.2 / 1M tokens $1.5 / 1M tokens
Together
Together | qwen/qwen3-coder-next-2025-02-03 262K $0.5 / 1M tokens $1.2 / 1M tokens
Chutes
Chutes | qwen/qwen3-coder-next-2025-02-03 262K $0.07 / 1M tokens $0.3 / 1M tokens
Parasail
Parasail | qwen/qwen3-coder-next-2025-02-03 262K $0.07 / 1M tokens $0.3 / 1M tokens
Chutes
Chutes | qwen/qwen3-coder-next-2025-02-03 262K $0.07 / 1M tokens $0.3 / 1M tokens
Parasail
Parasail | qwen/qwen3-coder-next-2025-02-03 262K $0.15 / 1M tokens $0.8 / 1M tokens
Benchmark Results
Benchmark Category Reasoning Strategy Free Executions Accuracy Cost Duration
Other Models by qwen