Qwen 3 Coder Next
Qwen3 Coder Next is an open-weight ultra-sparse Mixture-of-Experts model with 80 billion total parameters, activating just 3 billion per forward pass. It delivers roughly 10x higher throughput than comparable dense models, with a 256K context window extendable further for repository tasks. Released in February 2026 under the Apache 2.0 license, it scores 70.6 on SWE-Bench Verified and reports leading secure-code-generation results, outperforming a top proprietary model on the SecCodeBench security benchmark. It supports 370 programming languages with tool use and structured output for agentic coding. Open availability makes it practical to run frontier-grade local coding agents on-device, without cloud API dependency.
Key info
Benchmarks
Independent benchmark scores — composite indices for reasoning, coding, and math, plus individual eval scores where available.
Available routes
No routes currently available — Qwen 3 Coder Next isn't routed through the Opper gateway right now. It may return.
Contact us about this model →