Qwen3.8 Max Preview

Tool callingReasoningImage inputVideo input

Website

Preview Qwen flagship for million-token multimodal reasoning and long-horizon agentic workflows

Context: 1M
Released: 2mo ago

Sources:models.dev

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

No benchmark results have been matched to this model yet.

Specifications

Developer: Alibaba
Model family: Qwen
Type: Language model
Predecessor: —
Successor: —
Released: July 19, 2026
Model ID: alibaba/qwen3.8-max-preview
Openness: Closed
License: —
Input modalities: text, image, video
Output modalities: text
Modality: text+image+video->text
Max output tokens: 131K
Hugging Face: —
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Additional source evidence

Source model: alibaba/qwen3.8-max-preview
Source family: qwen
Last updated: 2026-07-19
Model context: 1M
Model output limit: 131K
Open weights: No
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Temperature control: Yes

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
Terminal-Bench	86.6 · accuracy	xhigh · 2.1 · 2026-08-03	Source
SWE-Bench Pro	67.7 · resolve rate	Claude Code · xhigh · 2026-08-03	Source
DeepSWE	56.6 · resolve rate	Claude Code · xhigh · 1.1 · 2026-08-03	Source
NL2Repo	55.9 · resolve rate	Claude Code · xhigh · 2026-08-03	Source
FrontierSWE	73.5 · dominance score	Claude Code · xhigh · 2026-08-03	Source
MLS-Bench-Lite	41 · score	Claude Code · xhigh · 2026-08-03	Source
AutomationBench	27.3 · pass@1	600-task public subset · xhigh · 2026-08-03	Source
Toolathlon Verified	72.5 · pass@1	xhigh · 2026-08-03	Source
WideSearch	81.9 · F1	xhigh · 2026-08-03	Source
Humanity's Last Exam	56.2 · accuracy	xhigh, with tools · 2026-08-03	Source
GPQA Diamond	92.6 · accuracy	xhigh · 2026-08-03	Source
Humanity's Last Exam	43.6 · accuracy	xhigh, no tools · 2026-08-03	Source
IFBench	82.8 · score	xhigh · 2026-08-03	Source
OSWorld-Verified	86.1 · success rate	xhigh · 2026-08-03	Source
MMMU Pro	82.3 · accuracy	xhigh · 2026-08-03	Source

Complete source record (JSON)

{
  "id": "alibaba/qwen3.8-max-preview",
  "name": "Qwen3.8 Max Preview",
  "description": "Preview Qwen flagship for million-token multimodal reasoning and long-horizon agentic workflows",
  "family": "qwen",
  "last_updated": "2026-07-19",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image",
      "video"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1000000,
    "output": 131072
  },
  "benchmarks": [
    {
      "name": "Terminal-Bench",
      "score": 86.6,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "accuracy",
      "variant": "xhigh",
      "version": "2.1",
      "date": "2026-08-03"
    },
    {
      "name": "SWE-Bench Pro",
      "score": 67.7,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "resolve rate",
      "harness": "Claude Code",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "DeepSWE",
      "score": 56.6,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "resolve rate",
      "harness": "Claude Code",
      "variant": "xhigh",
      "version": "1.1",
      "date": "2026-08-03"
    },
    {
      "name": "NL2Repo",
      "score": 55.9,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "resolve rate",
      "harness": "Claude Code",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "FrontierSWE",
      "score": 73.5,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "dominance score",
      "harness": "Claude Code",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "MLS-Bench-Lite",
      "score": 41,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "score",
      "harness": "Claude Code",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "AutomationBench",
      "score": 27.3,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "pass@1",
      "variant": "xhigh",
      "dataset": "600-task public subset",
      "date": "2026-08-03"
    },
    {
      "name": "Toolathlon Verified",
      "score": 72.5,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "pass@1",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "WideSearch",
      "score": 81.9,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "F1",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "Humanity's Last Exam",
      "score": 56.2,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "accuracy",
      "variant": "xhigh, with tools",
      "date": "2026-08-03"
    },
    {
      "name": "GPQA Diamond",
      "score": 92.6,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "accuracy",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "Humanity's Last Exam",
      "score": 43.6,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "accuracy",
      "variant": "xhigh, no tools",
      "date": "2026-08-03"
    },
    {
      "name": "IFBench",
      "score": 82.8,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "score",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "OSWorld-Verified",
      "score": 86.1,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "success rate",
      "variant": "xhigh",
      "date": "2026-08-03"
    },
    {
      "name": "MMMU Pro",
      "score": 82.3,
      "source": "https://www.alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421",
      "metric": "accuracy",
      "variant": "xhigh",
      "date": "2026-08-03"
    }
  ],
  "release_date": "2026-07-19",
  "open_weights": false
}

Infrastructure

Provider availability

All providers

Provider endpoint data will appear after this model's first OpenRouter endpoint sync.