Qwen3.8-MaxvsKimi K3
Across 11 shared benchmarks, Kimi K3 leads overall: Qwen3.8-Max wins 3, Kimi K3 wins 8, with 0 ties and an average score difference of -2.48.
Qwen3.8-Max
阿里巴巴 · 2026-08-03 · Reasoning model
Kimi K3
Moonshot AI · 2026-07-16 · Reasoning model
Qwen3.8-Max3 wins(27%)(73%)8 winsKimi K3
Benchmark scores
Grouped by capability, sorted by largest gap within each. 11 shared benchmarks.
AI Agent - Tool Usage
Kimi K3 3/3| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| Toolathlon-Verified | 72.504 / 5极高强度思考（工具） | 76.501 / 5Max (With Tools) | -4 |
| AutomationBench | 27.305 / 8极高强度思考（工具） | 30.803 / 8Max (With Tools) | -3.50 |
| Terminal-Bench 2.1 | 86.608 / 44极高强度思考（工具） | 88.302 / 44Max (With Tools) | -1.70 |
Coding and Software Engineer
Kimi K3 3/3| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| DeepSWE | 56.6014 / 27极高强度思考（工具） | 67.505 / 27Max (With Tools) | -10.90 |
| FrontierSWE | 73.504 / 4极高强度思考（工具） | 81.201 / 4Max (With Tools) | -7.70 |
| MLS Bench | 412 / 4极高强度思考（工具） | 48.301 / 4Max (With Tools) | -7.30 |
Math and Reasoning
Qwen3.8-Max 2/2| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| FrontierMath Tier 4 v2 | 46.3411 / 34极高强度思考（无工具） | 39.0213 / 34最高（无工具） | +7.32 |
| FrontierMath v2 | 74.7411 / 34极高强度思考（无工具） | 72.1813 / 34最高（无工具） | +2.55 |
Agent Level Benchmark
Kimi K3 1/1| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| Agents' Last Exam | 277 / 11极高强度思考（工具） | 28.305 / 11Max (With Tools) | -1.30 |
General Evaluation
Kimi K3 1/1| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| GPQA Diamond | 92.6021 / 226极高强度思考（无工具） | 93.5013 / 226最高（无工具） | -0.90 |
General Knowledge
Qwen3.8-Max 1/1| Benchmark | Qwen3.8-Max | Kimi K3 | Diff |
|---|---|---|---|
| HLE | 56.2013 / 181极高强度思考（工具） | 5614 / 181Max (With Tools) | +0.20 |
Specs
| Field | Qwen3.8-Max | Kimi K3 |
|---|---|---|
| Publisher | 阿里巴巴 | Moonshot AI |
| Release date | 2026-08-03 | 2026-07-16 |
| Model type | Reasoning model | Reasoning model |
| Architecture | MoE | MoE |
| Parameters | 2.4T | 2.8T |
| Context length | 1M | 1M |
| Max output | 128K | 1M |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Qwen3.8-Max | Kimi K3 |
|---|---|---|
| Text input | ¥12 / 1M tokens | ¥20 / 1M tokens |
| Text output | ¥36 / 1M tokens | ¥100 / 1M tokens |
| Cache read | ¥1.5 / 1M tokens | ¥2 / 1M tokens |
Summary
- Qwen3.8-Maxleads in:Math and Reasoning (2/2), General Knowledge (1/1)
- Kimi K3leads in:AI Agent - Tool Usage (3/3), Coding and Software Engineer (3/3), Agent Level Benchmark (1/1), General Evaluation (1/1)
On average across the 11 shared benchmarks, Kimi K3 scores 2.48 higher.
Largest single-benchmark gap: DeepSWE — Qwen3.8-Max 56.60 vs Kimi K3 67.50 (-10.90).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.