Skip to content
aiqbench

Alibaba · model passport

Qwen: Qwen3.5 397B A17B

qwen3-5-397b-a17b
  • Open weights
  • Apache 2.0 licence
  • 403.4B params
  • 262,144 context
  • Released 2026-02-16(HF repo created)
10/27axes covered
Overall · LMArena
1,442#84 of 396
Best percentile
96.9IFBench (Artificial Analysis)
Cohort standing
66.9median 56.8

What the data says

Benchmarked on 10 of 27 tracked axes. Every figure below is the model's own published result on that board.

Strongest axes

  • 96.9 pct — IFBench (Artificial Analysis) (78.8%)

Uneven: 57.3 percentile points between its best and worst axis — 96.9 pct on IFBench (Artificial Analysis) versus 39.6 pct on WebDev Arena.

  • IFBench (Artificial Analysis)96.9 pct78.8%Verify →
  • Hard prompts79.8 pct1,463Verify →
  • Instruction following79.5 pct1,433Verify →
  • Overall79 pct1,442Verify →
  • Coding78.8 pct1,491Verify →
  • GPQA Diamond71.5 pct86.4%Verify →
  • AIME (math)70.5 pct88.9%Verify →
  • Vision Arena66 pct1,246Verify →
  • Factuality Arena58.1 pct1,445Verify →
  • WebDev Arena39.6 pct1,400Verify →
  • Search ArenaNot benchmarked
  • Document ArenaNot benchmarked
  • Agent ArenaNot benchmarked
  • FrontierMathNot benchmarked
  • SimpleQA VerifiedNot benchmarked
  • ARC-AGI-2Not benchmarked
  • Humanity's Last ExamNot benchmarked
  • SimpleBenchNot benchmarked
  • LiveBenchNot benchmarked
  • BFCL v4 (function calling)Not benchmarked
  • SWE-bench Verified (live board)Not benchmarked
  • Terminal-Bench 4.0Not benchmarked
  • MMLU-Pro (Artificial Analysis)Not benchmarked
  • LiveCodeBench (Artificial Analysis)Not benchmarked
  • HHEM Hallucination (Vectara)Not benchmarked
  • SWE-bench VerifiedNot benchmarked
  • Terminal-BenchNot benchmarked

Percentiles are the model's standing within each board's own population. Axes marked "not benchmarked" have no published result for this model — that is an absence of data, not a low score.

How to run it

Run it yourself

RepositoryQwen/Qwen3.5-397B-A17B
Parameters403.4B
ArchitectureQwen3_5MoeForConditionalGeneration
LicenceApache 2.0

Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Qwen/Qwen3.5-397B-A17B"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

Ollama

We don't sync Ollama tags, so we won't print a command we can't verify — a wrong one would pull the wrong model. Search the Ollama library →

Unsloth fine-tuning recipes are model-specific and we don't derive them here. Unsloth docs →

Use it hosted

API model IDqwen/qwen3.5-397b-a17b
Context length262,144 tokens
Max output235,929 tokens
Acceptstext, image, video
Input price$0.50 per 1M tokens
Cached input price$0.30 per 1M tokens
Output price$3.60 per 1M tokens

API parameters this endpoint supports

Frequency penaltyInclude reasoningLogit biasLogprobsMax tokensMin-pPresence penaltyReasoningRepetition penaltyResponse formatSeedStop sequencesStructured outputsTemperatureTool choiceTool callingTop-kTop logprobsTop-p

Example request

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen/qwen3.5-397b-a17b","messages":[{"role":"user","content":"Hello"}]}'

View on OpenRouter →

Identifiers, limits and prices are synced from the upstream source shown above. Any field we haven't synced is left out entirely rather than guessed.

  • qwen3.5-397b-a17b stands at 66.9 (raw mean 72) — measured on 10 of 27 axes.

    standingtemplateVerify →

  • qwen3.5-397b-a17b stands at 66.9 versus a cohort median of 56.8 across 109 models released within 180 days — above the peer median.

    cohorttemplateVerify →

Every claim is a computed fact from the current board data, phrased by a low-temperature model call and rejected by a validation gate if any number is not in the source row. How the leaderboards work →