MAI-Thinking-1 vs Qwen3-Coder-480B-A35B-Instruct.
Verdict by category, tale of the tape, shared benchmarks only, and cost at three workloads — all from the tracked data.
Verdict by category
| Category | MAI-Thinking-1 | Qwen3-Coder-480B-A35B-Instruct | Winner |
|---|---|---|---|
| League Score | 46.7 | 41.3 | MAI-Thinking-1 |
| Coding (SWE-bench %) | — | 69.6 | — |
| Reasoning (GPQA %) | 84.2 | — | — |
| Context (tokens) | 1000000 | 262144 | MAI-Thinking-1 |
| Speed (tok/s) | — | — | — |
Tale of the tape
| Spec | MAI-Thinking-1 | Qwen3-Coder-480B-A35B-Instruct |
|---|---|---|
| Maker | Microsoft AI | Alibaba Qwen |
| Released | Jun 2, 2026 | Jul 23, 2025 |
| League Score | 47 | 41 |
| Context | 1,000,000 tokens | 262,144 tokens |
| Max output | — | — |
| $/1M input | — | — |
| $/1M output | — | — |
| Weights | Closed | Open |
| Speed | — | — |
Head-to-head benchmarks (shared only)
| Benchmark | MAI-Thinking-1 | Qwen3-Coder-480B-A35B-Instruct |
|---|---|---|
| No shared benchmark measurements yet — shown as not measured, never interpolated. | ||
Cost at three workloads
| Workload | MAI-Thinking-1 | Qwen3-Coder-480B-A35B-Instruct |
|---|---|---|
| Light — 1M in / 250K out | — | — |
| Standard — 10M in / 2.5M out | — | — |
| Long-context — 50M in / 5M out | — | — |
The pick
Pick MAI-Thinking-1 if you want the higher overall League Score; you need the bigger context window.
Pick Qwen3-Coder-480B-A35B-Instruct if you want downloadable weights / self-hosting; coding is the priority.
All figures from the model database — sources and dates on each model page. Cost your own workload or compare a different pair.
Frequently asked questions
Which is better overall, MAI-Thinking-1 or Qwen3-Coder-480B-A35B-Instruct?
MAI-Thinking-1 leads on overall League Score, 47 to 41 — a gap of 5 points on a composite of sourced benchmarks. That is overall form, not every category: check the per-category verdicts before choosing.
Which has the bigger context window, MAI-Thinking-1 or Qwen3-Coder-480B-A35B-Instruct?
MAI-Thinking-1, with 1M of context against 262.1K for Qwen3-Coder-480B-A35B-Instruct.
Are MAI-Thinking-1 and Qwen3-Coder-480B-A35B-Instruct open source?
MAI-Thinking-1 is closed weights — API access only; Qwen3-Coder-480B-A35B-Instruct is open weights — it can be downloaded and self-hosted.