Qwen3-Coder 480B A35B Instruct

Leaderboards

Where Qwen3-Coder 480B A35B Instruct sits among the top 20 in each category.

By task

Scored from votes on each kind of prompt alone, on the same 1,000-point scale.

Test results

Every test Qwen3-Coder 480B A35B Instruct has a published result on, out of 100%, with its rank among the models that took it.

• SWE-Bench Verified

  #72 / 120

  Fixing real bugs in real GitHub projects, checked by the projects' own tests.

  69.6%

  Self-reported

• Terminal-Bench 2.0

  #50 / 54

  Getting real work done in a command-line terminal.

  37.5%

  Self-reported

• SWE-bench Multilingual

  #40 / 49

  A multilingual benchmark for issue resolving in software engineering that covers Java, TypeScript, JavaScript, Go, Rust, C, and C++. Contains 1,632 high-quality instances carefully annotated from 2,456 candidates by 68 expert annotators, designed to evaluate Large Language Models across diverse software ecosystems beyond Python.

  54.7%

  Self-reported

• TAU-bench Retail

  #8 / 25

  A benchmark for evaluating tool-agent-user interaction in retail environments. Tests language agents' ability to handle dynamic conversations with users while using domain-specific API tools and following policy guidelines. Evaluates agents on tasks like order cancellations, address changes, and order status checks through multi-turn conversations.

  77.5%

  Self-reported

• TAU-bench Airline

  #5 / 23

  Part of τ-bench (TAU-bench), a benchmark for Tool-Agent-User interaction in real-world domains. The airline domain evaluates language agents' ability to interact with users through dynamic conversations while following domain-specific rules and using API tools. Agents must handle airline-related tasks and policies reliably.

  60.0%

  Self-reported

• Aider-Polyglot

  #11 / 22

  A coding benchmark that evaluates LLMs on 225 challenging Exercism programming exercises across C++, Go, Java, JavaScript, Python, and Rust. Models receive two attempts to solve each problem, with test error feedback provided after the first attempt if it fails. The benchmark measures both initial problem-solving ability and capacity to edit code based on error feedback, providing an end-to-end evaluation of code generation and editing capabilities across multiple programming languages.

  61.8%

  Self-reported

• BFCL-v3

  #13 / 20

  Berkeley Function Calling Leaderboard v3 (BFCL-v3) is an advanced benchmark that evaluates large language models' function calling capabilities through multi-turn and multi-step interactions. It introduces extended conversational exchanges where models must retain contextual information across turns and execute multiple internal function calls for complex user requests. The benchmark includes 1000 test cases across domains like vehicle control, trading bots, travel booking, and file system management, using state-based evaluation to verify both system state changes and execution path correctness.

  68.7%

  Self-reported

• Multi-SWE-Bench

  #7 / 7

  A multilingual benchmark for issue resolving that evaluates Large Language Models' ability to resolve software issues across diverse programming ecosystems. Covers 7 programming languages (Java, TypeScript, JavaScript, Go, Rust, C, and C++) with 1,632 high-quality instances carefully annotated by 68 expert annotators. Addresses limitations of existing benchmarks that focus almost exclusively on Python.

  25.8%

  Self-reported

• MM-Mind2Web

  #3 / 3

  A multimodal web navigation benchmark comprising 2,000 open-ended tasks spanning 137 websites across 31 domains. Each task includes HTML documents paired with webpage screenshots, action sequences, and complex web interactions.

  55.8%

  Self-reported

• Spider

  #2 / 2

  A large-scale, complex and cross-domain semantic parsing and text-to-SQL dataset annotated by 11 college students. Contains 10,181 questions and 5,693 unique complex SQL queries on 200 databases with multiple tables, covering 138 different domains. Requires models to generalize to both new SQL queries and new database schemas, making it distinct from previous semantic parsing tasks that use single databases.

  31.1%

  Unverified