Alibaba Cloud / Qwen Team model product
Qwen3.8 Max
4T-A95B open-weight checkpoint, is Qwen's flagship mixture-of-experts model for coding, research, professional workflows, and long-horizon agents.
Updated Aug 17, 2026. Default version: Qwen3.8 Max
LLMBoard Score89.8Qwen3.8 Max
Coverage80%33 benchmark families
Context window1MTokens
Official input price$2Alibaba API
Qwen3.8 Max Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Qwen3.8 Max Benchmark Results
Benchmark scores for Qwen3.8 Max.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkAndroidBench | Score75.1% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkAndroidWorld | Score85.3% | Rank01 | Participants5 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkCoWorkBench | Score74.8% | Rank01 | Participants4 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkERQA | Score77.8% | Rank01 | Participants24 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkHealthBench | Score60.2% | Rank01 | Participants9 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkIFBench | Score82.8% | Rank01 | Participants34 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkLongBench v2 | Score66.3% | Rank01 | Participants17 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkMobileWorld | Score77.8% | Rank01 | Participants3 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkOSWorld-Verified | Score86.1% | Rank01 | Participants24 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkPaperBench | Score93.0% | Rank01 | Participants3 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkPerceptionBench | Score63.5% | Rank01 | Participants2 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkPLawBench | Score73.2% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkPRBench-Finance | Score58.3% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkPRBench-Legal | Score57.6% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkQwenQoderBench | Score58.4% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkQwenReactBench | Score1,724 points | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkQwenSVG | Score1,713 points | Rank01 | Participants2 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkQwenSWEBench | Score80.7% | Rank01 | Participants2 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkRealWorldQA | Score88.0% | Rank01 | Participants29 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkSkillsBench | Score70.2% | Rank01 | Participants8 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkVideoMME w sub. | Score90.4% | Rank01 | Participants10 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkVision2Web | Score69.0% | Rank01 | Participants3 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkWideSearch | Score81.9% | Rank01 | Participants10 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkWorkspace Bench | Score67.7% | Rank01 | Participants3 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkAgents' Last Exam | Score52.4% | Rank02 | Participants10 | Percentile88.9% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam (with tools, text-only) | Score56.2% | Rank02 | Participants3 | Percentile50.0% | EvidenceC | Evaluated |
| BenchmarkJob Bench | Score53.4% | Rank02 | Participants4 | Percentile66.7% | EvidenceC | Evaluated |
| BenchmarkLVBench | Score81.8% | Rank02 | Participants25 | Percentile95.8% | EvidenceC | Evaluated |
| BenchmarkMLS-Bench Lite | Score41.0% | Rank02 | Participants3 | Percentile50.0% | EvidenceC | Evaluated |
| BenchmarkMRCR v2 (8-needle) | Score92.9% | Rank02 | Participants23 | Percentile95.5% | EvidenceC | Evaluated |
| BenchmarkNL2Repo | Score55.9% | Rank03 | Participants17 | Percentile87.5% | EvidenceC | Evaluated |
| BenchmarkOneMillion Bench | Score52.5% | Rank03 | Participants3 | Percentile0.0% | EvidenceC | Evaluated |
| BenchmarkScreenSpot Pro | Score84.5% | Rank03 | Participants25 | Percentile91.7% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Pro | Score67.7% | Rank04 | Participants50 | Percentile93.9% | EvidenceC | Evaluated |
| BenchmarkAutomationBench | Score27.3% | Rank05 | Participants12 | Percentile63.6% | EvidenceC | Evaluated |
| BenchmarkMMMU-Pro | Score82.3% | Rank05 | Participants68 | Percentile94.0% | EvidenceC | Evaluated |
| BenchmarkToolathlon | Score72.5% | Rank05 | Participants37 | Percentile88.9% | EvidenceC | Evaluated |
| BenchmarkFrontierSWE | Score73.5% | Rank06 | Participants16 | Percentile66.7% | EvidenceC | Evaluated |
| BenchmarkTerminal-Bench 2.1 | Score86.6% | Rank06 | Participants28 | Percentile81.5% | EvidenceC | Evaluated |
| BenchmarkGPQA | Score92.6% | Rank12 | Participants239 | Percentile95.4% | EvidenceC | Evaluated |
| BenchmarkDeepSWE 1.1 | Score56.6% | Rank13 | Participants25 | Percentile50.0% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam | Score43.6% | Rank30 | Participants99 | Percentile70.4% | EvidenceC | Evaluated |
Qwen3.8 Max Arena Results
Preference and agent-evaluation results for the default version.
Sort by
Arena | Category | Rank | Rating / score | Votes | Observations | Result date |
|---|
| Arenavision style control | Categoryoverall | Rank02 | Rating / score1301.3 | Votes5,344 | ObservationsN/A | Result date |
| Arenawebdev | Categoryoverall | Rank03 | Rating / score1666.7 | Votes2,855 | ObservationsN/A | Result date |
| Arenaagent task outcome explicit | Categoryoverall | Rank04 | Rating / score0.1 | VotesN/A | Observations5.4K | Result date |
| Arenatext | Categoryoverall | Rank06 | Rating / score1491.7 | Votes6,908 | ObservationsN/A | Result date |
| Arenavision | Categoryoverall | Rank06 | Rating / score1314.6 | Votes5,344 | ObservationsN/A | Result date |
| Arenatext style control | Categoryoverall | Rank09 | Rating / score1490.7 | Votes6,908 | ObservationsN/A | Result date |
| Arenaagent | Categoryoverall | Rank11 | Rating / score0.1 | VotesN/A | Observations561.8K | Result date |
| Arenaagent praise complaint | Categoryoverall | Rank11 | Rating / score0.1 | VotesN/A | Observations2K | Result date |
| Arenatext factuality | Categoryoverall | Rank11 | Rating / score1480.6 | Votes6,908 | ObservationsN/A | Result date |
| Arenaagent steerability | Categoryoverall | Rank18 | Rating / score0.1 | VotesN/A | Observations6.8K | Result date |
| Arenaagent bash recovery steps | Categoryoverall | Rank19 | Rating / score0.1 | VotesN/A | Observations14.5K | Result date |
| Arenaagent tool hallucination | Categoryoverall | Rank39 | Rating / score0.0 | VotesN/A | Observations533.2K | Result date |
Qwen3.8 Max Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- $2 input, $6 output per 1M
- Official provider
- Alibaba
- Lowest third-party
- From $1.65 input, $4.95 output per 1M via Deep Infra
- Tracked offerings
- 22
Sort by
Provider | Provider model ID | Region | Input / 1M | Output / 1M | Context | Updated |
|---|
| ProviderAlibaba Token Plan | Provider model IDqwen3.8-max | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderAlibaba Token Plan (China) | Provider model IDqwen3.8-max | Regionglobal | Input / 1MN/A | Output / 1MN/A | Context1M | Updated |
| ProviderDeep Infra | Provider model IDQwen/Qwen3.8-Max | Regionglobal | Input / 1M$1.65 | Output / 1M$4.95 | Context256K | Updated |
| ProviderAIHubMix | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$1.69 | Output / 1M$5.07 | Context991K | Updated |
| ProviderAlibaba (China) | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$1.78 | Output / 1M$5.33 | Context1M | Updated |
| ProviderLLM Gateway | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$1.82 | Output / 1M$5.45 | Context1M | Updated |
| ProviderCrossModel | Provider model IDqwen/qwen3.8-max | Regionglobal | Input / 1M$1.88 | Output / 1M$5.63 | Context1M | Updated |
| ProviderNanoGPT | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context991K | Updated |
| ProviderOpenRouter | Provider model IDqwen/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderOpenCode Go | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderMerge Gateway | Provider model IDqwen/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderDigitalOcean | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context262.1K | Updated |
| ProviderAlibaba | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderOfox | Provider model IDbailian/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderCharm Hyper | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderEmpirioLabs AI | Provider model IDqwen3-8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderVercel AI Gateway | Provider model IDalibaba/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderEden AI | Provider model IDqwen/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderKilo Gateway | Provider model IDqwen/qwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderFireworks AI | Provider model IDaccounts/fireworks/models/qwen3p8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context262.1K | Updated |
| ProviderAbacus | Provider model IDqwen3.8-max | Regionglobal | Input / 1M$2 | Output / 1M$6 | Context1M | Updated |
| ProviderVenice AI | Provider model IDqwen-3-8-max | Regionglobal | Input / 1M$2.5 | Output / 1M$7.5 | Context1M | Updated |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Qwen3.8 Max Runtime Performance
Provider-specific output speed and catalog latency for Qwen3.8 Max. Runtime does not affect the capability score.
No runtime data
Browse runtime rankingsNo provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Qwen3.8 Max Specifications
Technical details for the model's default version.
- Version
- Qwen3.8 Max
- Released
- Aug 2, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 2.4T
- Context window
- 1M
- Max output
- 131.1K
- Inputs
- image, pdf, text, video
- Outputs
- text
- Open weights
- No
- License
- Qwen3.8-Max License
Qwen3.8 Max Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionQwen3.8 Max | Released | LLMBoard89.8 | Parameters2.4T | Context1M | Max output131.1K | Open weightsNo | LicenseQwen3.8-Max License |
Qwen3.8 Max vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to Qwen3.8 Max
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
AC
Qwen3.7 Max
Alibaba Cloud / Qwen Team
AC
Qwen3.8 27B
Alibaba Cloud / Qwen Team
AC
Qwen3.7 Plus
Alibaba Cloud / Qwen Team
AC
Qwen3.6 Plus
Alibaba Cloud / Qwen Team
AC
Qwen3.5 397B A17B
Alibaba Cloud / Qwen Team
AC
Qwen3.6 27B
Alibaba Cloud / Qwen Team
What is Qwen3.8 Max?
Key information about Qwen3.8 Max and its available data.
4T-A95B open-weight checkpoint, is Qwen's flagship mixture-of-experts model for coding, research, professional workflows, and long-horizon agents. 4 trillion total parameters with 95 billion active parameters, a native 262,144-token context window extensible to about 1 million tokens, and always-on thinking.
8 Max API family with text and image input.
Data as of 2026-08-17.
FAQ
Common questions about Qwen3.8 Max.
When was Qwen3.8 Max released?
Qwen3.8 Max's default version was released on Aug 2, 2026.
How much does Qwen3.8 Max cost?
Qwen3.8 Max's official API price is $2 per million input tokens and $6 per million output tokens via Alibaba. The lowest tracked third-party offer starts at $1.65 input and $4.95 output via Deep Infra.
Who created Qwen3.8 Max?
Qwen3.8 Max was created by Alibaba Cloud / Qwen Team.
What is the context window for Qwen3.8 Max?
The default version has a 1M token context window.
Is Qwen3.8 Max open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer Qwen3.8 Max?
22 provider offerings are linked to the default version.
What models should I compare Qwen3.8 Max with?
Nearby ranked alternatives include Kimi K3, GLM 5.3, DeepSeek-V4-Pro.