modelbench日本語

Qwen3.8-Max

Alibaba(Qwen)Current

Input price
$2.00
Per 1 million tokens (USD)
Output price
$6.00
Per 1 million tokens (USD)
Effort levels
3total
Default: Xhigh
Benchmark results
15total
Benchmark types: 15

Model information

API model ID
qwen3.8-max
release date
2026-08-03
effort
Low / Med / Xhigh (default)
Effort source
Based on API Usage in the Alibaba Cloud announcement (alibabacloud.com/blog/qwen3-8-max-a-new-bar-for-coding-and-cowork_603421): reasoning_effort is xhigh (default), medium, low. In the Codex configuration example in the same article, default_reasoning_level is also xhigh and supported_reasoning_levels are low, medium, xhigh.
Claude Code
unavailable
Codex
unavailable
Availability notes
Official statements about Claude Code / Codex support have not been confirmed. The official model list for Claude Code / Codex (checked 2026-11-12) contains only Anthropic / OpenAI models, respectively, so both are false.
Availability source
https://www.alibabagroup.com/document-2021044032125272064

Price (USD per 1 million tokens)

Prices do not vary by effort; token consumption does.

input$2.00
output$6.00
cache read—
cache write—

Singapore (International). A single tier for input 0<Token≤1M. Both Thinking / Non-Thinking supported. snapshot qwen3.8-max-0902 has the same price. cache unit price not collected

qwen3.8-max | Non-Thinking and Thinking | 0<Token≤1M | $2 | $6

Source: https://www.alibabacloud.com/help/en/model-studio/model-pricing

Benchmark results

BenchmarkeffortScoreSource cost (metric)measured/reported bysourceConditions and extraction method
Terminal-Bench 2.1not statednot stated86.6%—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Claude Code (avg@10); 5 hour timeout; max_tokens=131,072.transcribed from table
Source excerptTerminal Bench 2.1 | 86.6Location HF model card "Benchmark Results" table, Terminal Bench 2.1 row, Qwen3.8-Max column (rightmost). Basis: note below the table 2
DeepSWE v1.1not statednot stated56.6%—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Higher of Claude Code and mini-SWE-agent (Claude Code for Qwen3.8-Max).transcribed from table
Source excerptDeepSWE 1.1 | 56.6Location HF model card "Benchmark Results" table, DeepSWE 1.1 row, Qwen3.8-Max column (rightmost). Basis: note below the table 4
GPQA DiamondXhigh (default)Xhigh92.7%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, gpqa_diamond.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptgpqa_diamond.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.9267676767676768, stderr=0.01686498021450146, Started at=2026-08-04T16:28:46.000ZLocation gpqa_diamond.csv, data row 145 (file line 205), id=TEVNDC9QXBh7PQvX9kHZip, Model version=qwen3.8-max_xhigh
FrontierMath Tiers 1-3(v2)Xhigh (default)Xhigh74.7%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tiers_1_3_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptfrontiermath_tiers_1_3_v2.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.7473684210526316, stderr=0.0257841025556124, Started at=2026-08-04T19:32:51.000ZLocation frontiermath_tiers_1_3_v2.csv, data row 82 (file line 108), id=ABT89ieSWGnGj4RYMduHiG, Model version=qwen3.8-max_xhigh
FrontierMath Tier 4(v2)Xhigh (default)Xhigh46.3%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, frontiermath_tier_4_v2.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptfrontiermath_tier_4_v2.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.4634146341463415, stderr=0.07884502378168105, Started at=2026-08-04T19:32:51.000ZLocation frontiermath_tier_4_v2.csv, data row 36 (file line 69), id=PBRdMZyz4L4CqPpXyLD2Lk, Model version=qwen3.8-max_xhigh
OTIS Mock AIME 2024-2025Xhigh (default)Xhigh99.4%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, otis_mock_aime_2024_2025.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptotis_mock_aime_2024_2025.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.9944444444444445, stderr=0.0038834738634418433, Started at=2026-08-04T16:28:46.000ZLocation otis_mock_aime_2024_2025.csv, data row 152 (file line 217), id=aGRKHaq9uSawfYVpC565zU, Model version=qwen3.8-max_xhigh
SimpleQA VerifiedXhigh (default)Xhigh45.8%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, simpleqa_verified.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptsimpleqa_verified.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.458, stderr=0.015763390640483554, Started at=2026-08-27T19:30:55.000ZLocation simpleqa_verified.csv, data row 55 (file line 100), id=bfftd5EzUK9yAekafEofvK, Model version=qwen3.8-max_xhigh
Chess Puzzles(Epoch AI)Xhigh (default)Xhigh46.0%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, chess_puzzles.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptchess_puzzles.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.46, stderr=0.05009082659620331, Started at=2026-10-13T13:13:49.000ZLocation chess_puzzles.csv, data row 143 (file line 199), id=bq5y6Tp83KLtxC58UDfpyN, Model version=qwen3.8-max_xhigh
MirrorCodeXhigh (default)Xhigh52.2%—Epoch AIEpoch AI, Capabilities & benchmarking (benchmark_data.zip, mirrorcode.csv) CC BY 4.0third-party · Retrieved 2026-11-12Run by Epoch AI (third-party measurement). Source label "Qwen3.8 Max (xhigh)". effort is indicated by the suffix _xhigh in Epoch's Model version. Data retrieved 2026-11-12.transcribed from table
Source excerptmirrorcode.csv: Model version=qwen3.8-max_xhigh, Best score (across scorers)=0.5222222222222223, stderr=0.10894718232315695, Started at=2026-10-21T16:48:06.000ZLocation mirrorcode.csv, data row 24 (file line 42), id=RQRuLNXoCHzNMEbJCcUVVD, Model version=qwen3.8-max_xhigh
Epoch Capabilities Index(ECI)not statednot stated157.59—Epoch AIEpoch AI, epoch_capabilities_index/eci_scores.csv (CC BY 4.0)third-party · Retrieved 2026-11-12A per-model metric that combines results across multiple benchmarks and multiple efforts (not per effort). Calculated by Epoch AI.transcribed from table
Source excerpteci_scores.csv: Model=Qwen 3.8 Max, eci=157.59, eci_ci_low=155.42, eci_ci_high=159.96Location epoch_capabilities_index/eci_scores.csv, data row 29 (file line 30), Model=Qwen 3.8 Max
SWE-Bench Pronot statednot stated67.7%—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Claude Code harness; 256K. All baselines evaluated on a version with problematic tasks fixed.transcribed from table
Source excerptSWE-bench Pro | 67.7Location HF model card "Benchmark Results" table, SWE-bench Pro row, Qwen3.8-Max column (rightmost). Basis: note below the table 3
NL2Repo-Benchnot statednot stated55.9—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Claude Code harness.transcribed from table
Source excerptNL2Repo-Bench | 55.9Location HF model card "Benchmark Results" table, NL2Repo-Bench row, Qwen3.8-Max column (rightmost). Basis: note below the table 5
MLS-Bench-Litenot statednot stated41.0%—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Claude Code; 5 hour timeout.transcribed from table
Source excerptMLS-Bench-Lite | 41.0Location HF model card "Benchmark Results" table, MLS-Bench-Lite row, Qwen3.8-Max column (rightmost). Basis: note below the table 7
AutomationBench (600 public tasks)not statednot stated27.3%—Alibaba (Qwen)Qwen3.8-2.4T-A95B (Hugging Face model card, Qwen)official · Retrieved 2026-11-13Vendor-reported value (self-reported by Alibaba's Qwen team). Table column is Qwen3.8-Max (card: Qwen3.8-Max is the official version based on Qwen3.8-2.4T-A95B). Effort not stated (the default for reasoning_effort is xhigh). Subset of 600 public tasks.transcribed from table
Source excerptAutomation-Bench (Pass@1) | 27.3Location HF model card "Benchmark Results" table, Automation-Bench (Pass@1) row, Qwen3.8-Max column (rightmost). Basis: note below the table 16
τ²-bench banking_knowledge (v1.0.1)Xhigh (default)Xhigh55.2%—Sierraτ²-bench leaderboard submission qwen3-8-max_sierra_2026-08-04 (sierra-research/tau2-bench, MIT)third-party · Retrieved 2026-11-13τ²-bench leaderboard submission qwen3-8-max_sierra_2026-08-04 (submission_type standard, tau2_bench_version 1.0.1). Submitter "Sierra", model_organization "Qwen": Run by Sierra, which operates τ²-bench (third-party measurement). Verified on the leaderboard (trajectories available, prompts not modified, no questions omitted). The value is pass^1 (%). reasoning_effort "xhigh". User simulator gpt-5.2 (reasoning_effort low). Search configuration retrieval_config "alltools". Evaluation date 2026-08-03. Notes summary: 4 trial runs; seed 300.transcribed from table
Source excerptbanking_knowledge.pass_1=55.154639175257735Location In submissions/qwen3-8-max_sierra_2026-08-04/submission.json: results.banking_knowledge.pass_1 (line 17 of file). model_name "Qwen 3.8 Max". methodology.notes (line 36)

Related rankings

References