Skip to main content
Radar

Five or fewer confirmed AI changes, with original sources, on mornings when something changed.A free source-linked morning brief.

See the free Radar Brief

Qwen3.5 397B

EstablishedReleased Feb 16, 2026Open WeightNon-Reasoning128K context

Released Feb 16, 2026 see all recent releases

Decision reading
Qwen3.5 397B scores 58 out of 100 and ranks #91 of 230. This profile shows 38 source-displayable benchmark rows; its strongest eligible category is Multilingual at #5. API pricing is $0.6 input and $3.6 output per million tokens.

Data as of September 2, 2026 · How the score is built

Strongest published evidence

Multilingual ranks #5. A well-rounded choice across a range of tasks.

Validate before choosing

38 published rows leave some tracked benchmark slots empty. Agentic is its lowest eligible category at #78.

Decision snapshot

Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.

Capability

58/100

field median 59.1

#91 of 230 ranked models

Price

$0.60input / $3.60 output

input median $1

blended $2.10

Speed

88tok/s

field median 90 tok/s

First token 38.58 s

Context

128Ktokens

field median 256,000

Reported for this model; direct source link not stored

Capability shape

Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.

Qwen3.5 397B category percentile values

  • Agentic46th percentile
  • Coding61st percentile
  • ReasoningNot eligible
  • Knowledge13th percentile
  • MathNot eligible
  • Multilingual64th percentile
  • Multimodal43rd percentile
  • Instruction following57th percentile

The dashed outline is median of 6 nearest peers.

Eligible category ranks

  1. Agentic#78/143
  2. Coding#59/148
  3. ReasoningNot ranked
  4. Knowledge#50/57
  5. MathNot ranked
  6. Multilingual#5/12
  7. Multimodal#21/36
  8. Inst. Following#19/43
Top decileTop quartileMid-fieldNot eligible

What it costs to get this score

Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.

Explore all models

The chart opens on the current model. Scroll horizontally to inspect the full price axis.

Current modelQwen3.5 397B · 58.0 score · $2.10 blended per million tokens
405060708090$0.50$1$5$10$25↘ frontierQwen3.5 397B

Horizontal: blended price per million tokens, log scale · Vertical: public score

How much of this is verified

Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.

  1. Agentic13/13 verified
  2. Coding3/3 verified
  3. Reasoning2/2 verified
  4. Knowledge6/6 verified
  5. Math5/5 verified
  6. Multilingual2/2 verified
  7. Multimodal6/6 verified
  8. Inst. Following1/1 verified
Verified sourceProvisionalNot measured

Spec sheet

Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.

API model ID
Not published
Context window
128K
Maximum output
Not sourced yet
Knowledge cutoff
Not sourced yet
Input modalities
Not sourced yet
Output modalities
Not sourced yet
Parameters
Not sourced yet
Availability
Not sourced yet
Cloud regions
Not tracked yet
Lifecycle
Established
API capabilities
Tool calling, structured outputs, and batch support are not tracked yet
Prompt caching
Not documented in the pricing record
Self-host
Open weights available; hardware estimate not sourced
Rate limits
Not tracked yet

Deployment options

Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.

Published weights are available, but BenchLM does not yet have a sourced parameter and VRAM profile for this exact model. Hardware cost estimates stay unavailable until that sizing record is complete.

Estimate VRAM from known parameters

Category score record

Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.

Category scores, ranks, weighting, benchmark coverage, and evidence status
CategoryScore
AgenticRank #78 of 143Percentile 46thWeight 22%13 benchmarksVerified48.0
CodingRank #59 of 148Percentile 61stWeight 20%3 benchmarksVerified54.1
ReasoningRank Not rankedWeight 17%2 benchmarksVerified65.7
KnowledgeRank #50 of 57Percentile 13thWeight 12%6 benchmarksVerified53.5
MathRank Not rankedWeight 5%5 benchmarksVerified74.3
MultilingualRank #5 of 12Percentile 64thWeight 7%2 benchmarksVerified69.7
MultimodalRank #21 of 36Percentile 43rdWeight 12%6 benchmarksVerified60.9
Inst. FollowingRank #19 of 43Percentile 57thWeight 5%1 benchmarkVerified87.4

Benchmark ledger

Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.

Coding3 rows
Coding benchmark values, best verified comparison, weight, and source status
SWE-bench VerifiedSoftware Engineering Benchmark VerifiedScore76.2%Versus best verified row

Best verified: Claude Opus 5 · 96%

Gap19.8 behindWeightWeighted 16%
SWE-bench ProScore50.9%Versus best verified row

Best verified: Claude Fable 5.1 · 81.2%

Gap30.3 behindWeightWeighted 10%
LiveCodeBench v6Score83.6%Versus best verified row

Best verified: Sakana Fugu-Ultra · 93.2%

Gap9.6 behindWeightDisplay only
Agentic13 rows
Agentic benchmark values, best verified comparison, weight, and source status
Terminal-Bench 2.0Score52.5%Versus best verified row

Best verified: GPT-5.6 Sol · 91.9%

Gap39.4 behindWeightWeighted 38%
BrowseCompScore62%Versus best verified row

Best verified: GPT-5.6 Sol · 92.2%

Gap30.2 behindWeightWeighted 28%
Claw-EvalScore56.8%Versus best verified row

Best verified: Ornith-1.5-397B · 81.4%

Gap24.6 behindWeightDisplay only
Benchmark exact
QwenClawBenchScore51.8%Versus best verified row

Best verified: Qwen3.7 Max · 64.3%

Gap12.5 behindWeightDisplay only
τ³-bench resultsτ³-Bench Tool-Agent-User EvaluationScore68.4%Versus best verified row

Best verified: Mistral Medium 3.5 128B · 91.4%

Gap23 behindWeightDisplay only
VITA-BenchScore43.7%Versus best verified row

Best verified: Qwen3.7 Max · 47.9%

Gap4.2 behindWeightDisplay only
DeepPlanningScore37.6%Versus best verified row

Best verified: Qwen3.7 Plus · 62.3%

Gap24.7 behindWeightDisplay only
ToolathlonScore36.3%Versus best verified row

Best verified: Muse Spark 1.1 · 75.6%

Gap39.3 behindWeightDisplay only
MCP AtlasScore46.1%Versus best verified row

Best verified: Muse Spark 1.1 · 88.1%

Gap42 behindWeightDisplay only
MCP-TasksScore74.2%Versus best verified row

Best verified: Qwen3.5 397B · 74.2%

GapBest verifiedWeightDisplay only
WideResearchScore74.0%Versus best verified row

Best verified: Hy4 preview · 83.9%

Gap9.9 behindWeightDisplay only
Gert LabsGert Labs Composite Game BenchmarkScore46.76%Versus best verified row

Best verified: Claude Opus 4.8 · 72.97%

Gap26.2 behindWeightDisplay only
Benchmark exact
ResearchClawBenchScore14.2%Versus best verified row

Best verified: Claude Opus 4.8 · 21.1%

Gap6.9 behindWeightDisplay only
Reasoning2 rows
Reasoning benchmark values, best verified comparison, weight, and source status
LongBench v2Score63.2%Versus best verified row

Best verified: Qwen3.8 Max · 66.3%

Gap3.1 behindWeightWeighted 38%
AI-NeedleScore68.7%Versus best verified row

Best verified: Qwen3.5 397B · 68.7%

GapBest verifiedWeightDisplay only
Knowledge6 rows
Knowledge benchmark values, best verified comparison, weight, and source status
HLEHumanity's Last ExamScore28.7%Versus best verified row

Best verified: Claude Fable 5.1 · 65%

Gap36.3 behindWeightWeighted 45%
MMLU-ProMassive Multitask Language Understanding ProfessionalScore87.8%Versus best verified row

Best verified: Qwen3.7 Max · 89.6%

Gap1.8 behindWeightWeighted 30%
GPQAGraduate-Level Google-Proof Q&AScore88.4%Versus best verified row

Best verified: Sakana Fugu-Ultra · 95.5%

Gap7.1 behindWeightWeighted 7%
SuperGPQASuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesScore70.4%Versus best verified row

Best verified: Qwen 3.6 Max (preview) · 73.9%

Gap3.5 behindWeightWeighted 7%
MMLU-ReduxScore94.9%Versus best verified row

Best verified: Qwen3.7 Max · 95%

Gap0.1 behindWeightDisplay only
C-EvalScore93%Versus best verified row

Best verified: Qwen3.6 Plus · 93.3%

Gap0.3 behindWeightDisplay only
Math5 rows
Math benchmark values, best verified comparison, weight, and source status
AIME26AIME 2026Score93.3%Versus best verified row

Best verified: GLM-5.2 · 99.2%

Gap5.9 behindWeightWeighted 25%
HMMT Feb 2026Harvard-MIT Mathematics Tournament February 2026Score87.9%Versus best verified row

Best verified: Qwen3.7 Max · 97.1%

Gap9.2 behindWeightWeighted 25%
HMMT Feb 2025Harvard-MIT Mathematics Tournament February 2025Score94.8%Versus best verified row

Best verified: Qwen3.6 Plus · 96.7%

Gap1.9 behindWeightDisplay only
HMMT Nov 2025Harvard-MIT Mathematics Tournament November 2025Score92.7%Versus best verified row

Best verified: Qwen3.6 Plus · 94.6%

Gap1.9 behindWeightDisplay only
MMAnswerBenchScore80.9%Versus best verified row

Best verified: GLM-5.2 · 91.0%

Gap10.1 behindWeightDisplay only
Multilingual2 rows
Multilingual benchmark values, best verified comparison, weight, and source status
MMLU-ProXScore84.7%Versus best verified row

Best verified: Qwen3.7 Max · 87%

Gap2.3 behindWeightWeighted 100%
NOVA-63Score59.1%Versus best verified row

Best verified: Qwen3.5 397B · 59.1%

GapBest verifiedWeightDisplay only
Multimodal6 rows
Multimodal benchmark values, best verified comparison, weight, and source status
MMMU-ProMassive Multi-discipline Multimodal Understanding ProScore79%Versus best verified row

Best verified: GPT-5.4 Pro · 94%

Gap15 behindWeightWeighted 45%
CharXivCharXiv ReasoningScore80.8%Versus best verified row

Best verified: Qwen3.8 Max · 93.5%

Gap12.7 behindWeightWeighted 25%
MathVisionScore88.6%Versus best verified row

Best verified: Qwen3.8 Max · 95.2%

Gap6.6 behindWeightDisplay only
VideoMMMUScore84.7%Versus best verified row

Best verified: Qwen3.8 Max · 88.7%

Gap4 behindWeightDisplay only
ScreenSpot ProScore65.6%Versus best verified row

Best verified: Claude Opus 4.8 · 87.9%

Gap22.3 behindWeightDisplay only
V*Score95.8%Versus best verified row

Best verified: Kimi K2.6 · 96.9%

Gap1.1 behindWeightDisplay only
Inst. Following1 row
Inst. Following benchmark values, best verified comparison, weight, and source status
IFEvalInstruction-Following EvalScore92.6%Versus best verified row

Best verified: Qwen3.5-27B · 95%

Gap2.4 behindWeightWeighted 35%

Lineage

The sequence follows explicit supersedes links. A successor's displayed score stays at least 0.1 points above its predecessor; raw benchmark rows do not move. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.

  1. Feb 16, 2026 · you are here

    Qwen3.5 397B

    Score 58.0 · $0.6 / $3.6

Base entry

How to read this profile

The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.

Qwen3.5 397B ranks #91 of 230 on the public leaderboard with a score of 57.98/100. It does not yet have enough sourced coverage for a verified position.

Qwen3.5 397B is a open weight model with a 128K context window. No explicit reasoning mode is documented in this profile.

Qwen3.5 397B sits in the Qwen3.5 397B family with Qwen3.5 397B (Reasoning). 38 of 416 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.

Its strongest eligible category is Multilingual at #5, while its lowest eligible position is Agentic at #78. a well-rounded choice across a range of tasks.

Frequently asked questions

How does Qwen3.5 397B perform overall in AI benchmarks?

Qwen3.5 397B ranks #91 out of 230 models on the public BenchAlign leaderboard, with a score of 57.98/100. Its evidence status is Estimated, and this profile shows 38 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.

Is Qwen3.5 397B good for knowledge and understanding?

Qwen3.5 397B ranks #50 out of 57 eligible models for knowledge and understanding, with a public category score of 53.5/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B good for coding and programming?

Qwen3.5 397B ranks #59 out of 148 eligible models for coding and programming, with a public category score of 54.1/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B good for mathematics?

Qwen3.5 397B has source-displayable benchmark coverage for mathematics, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.

Is Qwen3.5 397B good for reasoning and logic?

Qwen3.5 397B has source-displayable benchmark coverage for reasoning and logic, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.

Is Qwen3.5 397B good for agentic tool use and computer tasks?

Qwen3.5 397B ranks #78 out of 143 eligible models for agentic tool use and computer tasks, with a public category score of 48/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B good for multimodal and grounded tasks?

Qwen3.5 397B ranks #21 out of 36 eligible models for multimodal and grounded tasks, with a public category score of 60.9/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B good for instruction following?

Qwen3.5 397B ranks #19 out of 43 eligible models for instruction following, with a public category score of 87.4/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B good for multilingual tasks?

Qwen3.5 397B ranks #5 out of 12 eligible models for multilingual tasks, with a public category score of 69.7/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.5 397B open source?

Qwen3.5 397B is an open-weight model from Alibaba. Its weights can be downloaded for local or hosted deployment, subject to the published license. Open weight does not automatically mean open source: training data and training code may remain private, and commercial restrictions can still apply.

Which sibling models are related to Qwen3.5 397B?

Qwen3.5 397B belongs to the Qwen3.5 397B family. Related tracked variants include Qwen3.5 397B (Reasoning). A sibling link indicates shared lineage or a documented configuration relationship; it does not mean the variants have identical pricing, context limits, benchmark evidence, or deployment behavior. Compare before switching.

Does Qwen3.5 397B have full benchmark coverage on BenchLM?

No. Qwen3.5 397B currently has 55 source-displayable rows across 416 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.

What is the context window size of Qwen3.5 397B?

Qwen3.5 397B has a reported context window of 128K in the exact-model catalog record. The value stays visible, but the profile marks its source link as unavailable instead of presenting it as directly documented. Maximum output length remains separate because providers often publish a different limit.

Compare Qwen3.5 397B with every tracked model409 comparisons

Last updated September 2, 2026. Runtime fields remain blank until a sourced snapshot exists.

Watch Qwen3.5 397B in the weekly brief

Get one weekly email when material rank, price, availability, or benchmark evidence changes are worth revisiting.

Read a sample issue

Join 2,000+ readers.