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.

Start free brief

Qwen3.8-Flash-Next

Released Aug 26, 2026 see all recent releases

Decision reading
Qwen3.8-Flash-Next scores 67.5 out of 100 and ranks #25 of 226. This profile shows 24 source-displayable benchmark rows; its strongest eligible category is Instruction Following at #5. Published weights can be self-hosted, but infrastructure cost varies and is not a comparable API token rate.

Data as of August 26, 2026 · How the score is built

Strongest published evidence

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

Validate before choosing

24 published rows leave some tracked benchmark slots empty. No comparable first-party API token rate is published.

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

67.5/100

field median 58.3

#25 of 226 ranked models

Price

Self-hosted; infrastructure cost varies

input median $1

No comparable first-party hosted token rate

Speed

Not measured

field median 97 tok/s

Time to first token not measured

Context

262Ktokens

field median 203,000

Maximum output length is tracked separately

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.

Qwen3.8-Flash-Next category percentile values

  • Agentic91st percentile
  • Coding78th percentile
  • ReasoningNot eligible
  • KnowledgeNot eligible
  • MathNot eligible
  • MultilingualNot eligible
  • MultimodalNot eligible
  • Instruction following90th percentile

The dashed outline is median of 6 nearest peers.

Eligible category ranks

  1. Agentic#13/138
  2. Coding#32/144
  3. ReasoningNot ranked
  4. KnowledgeNot ranked
  5. MathNot ranked
  6. MultilingualNot ranked
  7. MultimodalNot ranked
  8. Inst. Following#5/42
Top decileTop quartileMid-fieldNot eligible

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. Agentic6/6 verified
  2. Coding5/5 verified
  3. ReasoningNot measured
  4. Knowledge4/4 verified
  5. MathNot measured
  6. MultilingualNot measured
  7. Multimodal8/8 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
Qwen/Qwen3.8-Flash-NextQwen3.8-Flash-Next model card
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
Qwen publishes the BF16 post-trained checkpoint on Hugging Face under the Qwen Community 1.0 license. The model card documents native text, image, and video input, reasoning-effort controls, and a 262,144-token native context window that can be extended to 1,000,000 tokens with YaRN. Qwen Cloud offers a separate production Qwen3.8-Flash model based on this architecture; this row does not inherit that sibling's features or pricing.
Cloud regions
Not tracked yet
Lifecycle
Current
API capabilities
Tool calling, structured outputs, and batch support are not tracked yet
Prompt caching
Not documented in the pricing recordQwen3.8-Flash-Next model card
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 #13 of 138Percentile 91stWeight 22%6 benchmarksVerified62.7
CodingRank #32 of 144Percentile 78thWeight 20%5 benchmarksVerified61.2
ReasoningWeight 17%0 benchmarksNot measuredNot measured
KnowledgeRank Not rankedWeight 12%4 benchmarksVerified55.0
MathWeight 5%0 benchmarksNot measuredNot measured
MultilingualWeight 7%0 benchmarksNot measuredNot measured
MultimodalRank Not rankedWeight 12%8 benchmarksVerified81.8
Inst. FollowingRank #5 of 42Percentile 90thWeight 5%1 benchmarkVerified94.4

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.

Coding5 rows
Coding benchmark values, best verified comparison, weight, and source status
SWE-bench ProScore62.5%Versus best verified row

Best verified: Claude Mythos 5 · 80.3%

Gap17.8 behindWeightWeighted 10%
SWE MultilingualScore81%Versus best verified row

Best verified: Claude Opus 5 · 89.5%

Gap8.5 behindWeightDisplay only
NL2RepoScore48.1%Versus best verified row

Best verified: DeepSeek V4 Pro 0813 · 61.5%

Gap13.4 behindWeightDisplay only
deepSweScore58.7%Versus best verified row

Best verified: GPT-5.6 Sol · 72.7%

Gap14 behindWeightDisplay only
LiveCodeBench v6Score91.9%Versus best verified row

Best verified: Sakana Fugu-Ultra · 93.2%

Gap1.3 behindWeightDisplay only
Agentic6 rows
Agentic benchmark values, best verified comparison, weight, and source status
CoWorkBenchScore73.9%Versus best verified row

Best verified: Qwen3.8 Max · 74.8%

Gap0.9 behindWeightDisplay only
JobBenchScore55.7%Versus best verified row

Best verified: Qwen3.8-Flash-Next · 55.7%

GapBest verifiedWeightDisplay only
Agents' Last ExamScore51.2%Versus best verified row

Best verified: Qwen3.8 Max · 52.4%

Gap1.2 behindWeightDisplay only
Toolathlon-VerifiedScore73.5%Versus best verified row

Best verified: Claude Opus 5 · 80.6%

Gap7.1 behindWeightDisplay only
AndroidWorldScore84.5%Versus best verified row

Best verified: Qwen3.8 Max · 85.3%

Gap0.8 behindWeightDisplay only
OSWorld 2.0Score19.4%Versus best verified row

Best verified: Claude Opus 5 · 70.6%

Gap51.2 behindWeightDisplay only
Knowledge4 rows
Knowledge benchmark values, best verified comparison, weight, and source status
HLEHumanity's Last ExamScore35.9%Versus best verified row

Best verified: Claude Opus 5 · 64.7%

Gap28.8 behindWeightWeighted 45%
GPQAGraduate-Level Google-Proof Q&AScore91.7%Versus best verified row

Best verified: Sakana Fugu-Ultra · 95.5%

Gap3.8 behindWeightWeighted 7%
GPQA-DGPQA DiamondScore91.7%Versus best verified row

Best verified: Sakana Fugu-Ultra · 95.5%

Gap3.8 behindWeightDisplay only
HLE w/o toolsHumanity's Last Exam without toolsScore35.9%Versus best verified row

Best verified: Claude Mythos 5 · 59%

Gap23.1 behindWeightDisplay only
Multimodal8 rows
Multimodal benchmark values, best verified comparison, weight, and source status
CharXivCharXiv ReasoningScore90.6%Versus best verified row

Best verified: Claude Mythos 5 · 93.5%

Gap2.9 behindWeightWeighted 25%
Vision2WebScore64.0%Versus best verified row

Best verified: Qwen3.8 Max · 69.0%

Gap5 behindWeightDisplay only
ERQAScore72.3%Versus best verified row

Best verified: Qwen3.8 Max · 77.8%

Gap5.5 behindWeightDisplay only
LVBenchScore76.6%Versus best verified row

Best verified: Qwen3.8 Max · 81.8%

Gap5.2 behindWeightDisplay only
RealWorldQAScore88.5%Versus best verified row

Best verified: Qwen3.8-Flash-Next · 88.5%

GapBest verifiedWeightDisplay only
MathVisionScore90.6%Versus best verified row

Best verified: Qwen3.8 Max · 95.2%

Gap4.6 behindWeightDisplay only
MathVision w/ PythonMathVision with PythonScore95.7%Versus best verified row

Best verified: Kimi K3 · 97.8%

Gap2.1 behindWeightDisplay only
CharXiv w/o toolsCharXiv Reasoning without toolsScore84.6%Versus best verified row

Best verified: Claude Mythos 5 · 88.9%

Gap4.3 behindWeightDisplay only
Inst. Following1 row
Inst. Following benchmark values, best verified comparison, weight, and source status
IFBenchInstruction Following BenchmarkScore81.3%Versus best verified row

Best verified: MAI-Thinking-1 · 85%

Gap3.7 behindWeightWeighted 65%

Lineage

The sequence follows explicit supersedes links. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.

experimental-preview · Next

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.

Qwen3.8-Flash-Next ranks #25 of 226 on the public leaderboard with a score of 67.54/100. It does not yet have enough sourced coverage for a verified position.

Qwen3.8-Flash-Next is a open weight model with a 262K context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.

Qwen publishes the BF16 post-trained checkpoint on Hugging Face under the Qwen Community 1.0 license. The model card documents native text, image, and video input, reasoning-effort controls, and a 262,144-token native context window that can be extended to 1,000,000 tokens with YaRN. Qwen Cloud offers a separate production Qwen3.8-Flash model based on this architecture; this row does not inherit that sibling's features or pricing.

Official exact-value snapshot from Qwen's August 26, 2026 Qwen3.8-Flash-Next model card and technical report. We map only the post-trained model-card rows that match existing protocols, keep the 262,144-token native context instead of the optional 1M YaRN extension, and preserve paired no-Code-Interpreter and Code-Interpreter visual scores separately. DeepSWE 1.1 uses the best result across Claude Code and mini-SWE-agent, SWE-bench Pro uses Qwen's corrected-task Claude Code run, and the remaining harness-specific agent results stay display-only. ClawEval-MM, RecreationBench, and the technical report's 14 base-model benchmarks remain outside this post-trained row rather than being collapsed into non-equivalent or cross-stage fields.

24 of 406 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.

Its strongest eligible category is Instruction Following at #5, while its lowest eligible position is Coding at #32. a well-rounded choice across a range of tasks.

Radar

Qwen3.8-Flash-Next release history

Full release history

Frequently asked questions

How does Qwen3.8-Flash-Next perform overall in AI benchmarks?

Qwen3.8-Flash-Next ranks #25 out of 226 models on the public BenchAlign leaderboard, with a score of 67.54/100. Its evidence status is Estimated, and this profile shows 24 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.

Is Qwen3.8-Flash-Next good for knowledge and understanding?

Qwen3.8-Flash-Next has source-displayable benchmark coverage for knowledge and understanding, 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 Qwen3.8-Flash-Next good for coding and programming?

Qwen3.8-Flash-Next ranks #32 out of 144 eligible models for coding and programming, with a public category score of 61.2/100. 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 Qwen3.8-Flash-Next good for agentic tool use and computer tasks?

Qwen3.8-Flash-Next ranks #13 out of 138 eligible models for agentic tool use and computer tasks, with a public category score of 62.7/100. 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 Qwen3.8-Flash-Next good for multimodal and grounded tasks?

Qwen3.8-Flash-Next has source-displayable benchmark coverage for multimodal and grounded tasks, 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 Qwen3.8-Flash-Next good for instruction following?

Qwen3.8-Flash-Next ranks #5 out of 42 eligible models for instruction following, with a public category score of 94.4/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.

Is Qwen3.8-Flash-Next open source?

Qwen3.8-Flash-Next 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.

Does Qwen3.8-Flash-Next have full benchmark coverage on BenchLM?

No. Qwen3.8-Flash-Next currently has 24 source-displayable rows across 406 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 Qwen3.8-Flash-Next?

Qwen3.8-Flash-Next has a documented context window of 262K. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.

Compare Qwen3.8-Flash-Next with every tracked model396 comparisons

Last updated August 26, 2026. Runtime fields remain blank until a sourced snapshot exists.

Watch Qwen3.8-Flash-Next 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.