Qwen3.8 27B
Qwenfamily · Qwen3.8huggingface.co/Qwen/Qwen3.8-27B
Qwen3.8 27B is an open-weight dense vision-language model from Qwen. It is suited for coding, professional workflows, research, multimodal interaction, and long-running agent tasks, with flexible thinking that can be...
Updated 2 h ago · first seen 11 Sept 2026
model_01M294WVR4JWTTVTSEYX2E0VCE
Overview
Identity
- Canonical model
- Yesidentity confidence: mediumOne row per real model release. Artifacts (checkpoints, quantisations, conversions) and folded evaluation variants point here.
- Official checkpoints
- official_checkpoints = hf_repo identifiers carried by the model itself; artifacts are separate entities pointing here through canonical_id.
- Artifacts
- 9 quantizations0 official · 8 third-partySeparate entities (checkpoint · quantization · conversion · packaging) pointing to this model through canonical_id.
- Provider deployments
- 2
- API aliases
- qwen/qwen3.8-27bqwen3-8-27bqwen3-8-27b-lowqwen3-8-27b-mediumqwen3-8-27b-non-reasoningIdentifiers under which providers and evaluators refer to this model.
- Folded evaluation variants
- 3Effort / thinking variants (…-high, …-non-reasoning) are result configurations of this model, not separate models. Their old URLs redirect here.
Openness
Open weights— weights downloadable under Apache-2.0; commercial use allowed; redistribution allowed; derivatives allowed; 4 dimensions unknown.
Weights downloadable under a permissive or Creative Commons licence allowing commercial use; code or data may be missing.
Weights
Yes
Inference code
—
Training code
—
Training data
—
Dataset
—
Commercial use
Yes
Redistribution
Yes
Derivatives
Yes
Licence: Apache License 2.0 (permissive · SPDX Apache-2.0 · stated as “apache-2.0”)
dimensions marked null are unknown, not false
Key facts
- Release date
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 7 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Openrouter id
Source:OpenRouter public model & pricing listingT2observed 12 h agomedium
Architecture
- Architecture
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Model type
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Parameters
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Tokenizer
Source:OpenRouter public model & pricing listingT2observed 12 h agomedium
- Weights dtype
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- File size
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Library name
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 11 h agomedium
- Pipeline tag
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium
- Hugging Face repo
Source:OpenRouter public model & pricing listingT2observed 12 h agomedium
Capabilities
Modalities
- Modalities
- imagetextvideo
- Input
- imagetextvideo
- Output
- text
Capabilities
Tool calling
Yes
OpenRouter public model & pricing listing · T2
Structured output
Yes
OpenRouter public model & pricing listing · T2
Reasoning
Yes
OpenRouter public model & pricing listing · T2
Vision
Yes
OpenRouter public model & pricing listing · T2
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
Source:OpenRouter public model & pricing listingT2observed 9 h agomedium
- Max output
Source:OpenRouter public model & pricing listingT2observed 12 h agomedium
- Tokenizer
Source:OpenRouter public model & pricing listingT2observed 12 h agomedium
Benchmarks62
Compare with another model →Comparable same task and conditions · Partially comparable same task, conditions differ (effort, temperature, judge) · Not comparable different variant or metric
Current rows only, grouped by benchmark → canonical metric → comparability group (task configuration). Effort variants folded into this model appear as rows of the same group. 62 current rows in total. “vs leader” compares with the current leader of the benchmark's primary group only; other groups are not directly comparable. Comparability rules →
Providers & Pricing2
All offers in the price terminal →USD per 1M tokens as published by each provider; native units (per-request fees, flex/priority tiers) are kept verbatim. Rows are append-only — every price change is kept in the history below. Cost of a workload →
Price history
Output price · USD / 1M tokens 2 providers
- OpenRouter
- GroqCloud
- OpenRouter$3 → $2.5512 Sept 2026
- GroqCloudfirst observed $411 Sept 2026
- OpenRouterfirst observed $311 Sept 2026
Input price · USD / 1M tokens 2 providers
- OpenRouter
- GroqCloud
- OpenRouter$0.42 → $0.21412 Sept 2026
- GroqCloudfirst observed $0.8011 Sept 2026
- OpenRouterfirst observed $0.4211 Sept 2026
Hardware fit37
Assumptions (7)
- Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
- bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
- KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
- A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
- Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
- Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
- Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.
Lineage
Open in Graph →Versions & Artifacts9
Version history
Context window2 changes
11 Sept 2026→11 Sept 2026→12 Sept 2026current
Licensefirst observation only
11 Sept 2026current
Max outputfirst observation only
11 Sept 2026current
Opennessfirst observation only
11 Sept 2026current
Parametersfirst observation only
11 Sept 2026current
Each hop is a claim: click a value for its source, tier and observation time. Nothing is overwritten — a new observation closes the previous claim.
Artifacts 9
quantizations 9
- Qwen/Qwen3.8-27B-FP8Qwen · FP8 · BF16/F8_E4M330.9 GB
- amd/Qwen3.8-27B-Quark-AWQ-INT4-W4A16AMD · AWQ · BF16/I3219.5 GB
- amd/Qwen3.8-27B-Quark-AWQ-MXFP4AMD · AWQ · BF16/U819.8 GB
- bartowski/Qwen3.8-27B-GGUFbartowski · GGUF—
- mlx-community/Qwen3.8-27B-4bitMLX Community · MLX · BF16/U3216.1 GB
- mlx-community/Qwen3.8-27B-8bitMLX Community · MLX · BF16/U3229.5 GB
- perplexity-ai/pplx-computer-qwen-3-8-27b-dflash2-gguf-20260826Perplexity AI · GGUF—
- unsloth/Qwen3.8-27B-GGUFUnsloth · GGUF—
- unsloth/Qwen3.8-27B-NVFP4Unsloth · BF16/F8_E4M3/U822.6 GB
Timeline30
Full timeline →Qwen3.8 27B scores 12.14% on Humanity's Last Exam
artificial_analysisQwen3.8 27B scores 22.42 on Artificial Analysis Intelligence Index
artificial_analysisQwen3.8 27B scores 14.09% on Humanity's Last Exam
artificial_analysisQwen3.8 27B scores 27.81 on Artificial Analysis Intelligence Index
artificial_analysisQwen3.8 27B scores 14.04% on Humanity's Last Exam
artificial_analysisQwen3.8 27B scores 26.49 on Artificial Analysis Intelligence Index
artificial_analysisQwen3.8 27B scores 33.92% on Humanity's Last Exam
artificial_analysisQwen3.8 27B scores 33.9 on Artificial Analysis Intelligence Index
artificial_analysisOpenRouter changed pricing for Qwen3.8 27B: $0.42 in / $3 out per 1M tokens → $0.214 in / $2.55 out per 1M tokens
$0.42 in / $3 out→$0.214 in / $2.55 outopenrouterQwen3.8 27B: context length changed from 256000 to 1000000
Context window256K tokens→1M tokensopenrouterQwen3.8 27B: release date changed from 2026-08-05 to 2026-08-14
Release date5 Aug 2026→14 Aug 2026openrouter
Change history68
Viewing AI Atlas as of 12 Sept 2026 — attributes exactly as the atlas knew them on that day; later corrections are not shown.
Back to today →Attributes as of 12 Sept 2026 43 claims in force
- Release date
- 5 Aug 2026
- Openness
- open-weights
- License
- Apache-2.0
- Architecture
- Qwen3_5ForConditionalGeneration
- Parameters
- 27.8B
- Context window
- 1M tokens
- Max output
- 131.1K tokens
- Modalities
- image, text, video
- Input modalities
- image, text, video
- Output modalities
- text
- Tokenizer
- Qwen
- File size
- 55.6 GB
- Hugging Face repo
- Qwen/Qwen3.8-27B
- Pipeline tag
- image-text-to-text
- Model card
- Downloads
- 7,563,763
- Likes
- 14,766
- Aa context window
- 256,000
- Aa openness
- open-weights
- Commercial use allowed
- Yes
- Derivatives allowed
- Yes
- Gated
- No
- Groq model id
- qwen/qwen3.8-27b
- Groq status
- preview
- Hf inference providers
- cerebras, deepinfra, featherless-ai, novita, ovhcloud
- Last modified
- 2026-08-14T15:00:01+00:00
- Library name
- transformers
- Livebench hf link
- https://huggingface.co/Qwen/Qwen3.8-27B
- Aa median output tokens per second
- 39
- Downloads all time
- 7,563,763
- Model type
- qwen3_5
- Openrouter id
- qwen/qwen3.8-27b
- Openrouter listed at
- 14 Aug 2026
- Reasoning
- Yes
- Redistribution allowed
- Yes
- Structured output
- Yes
- Supported parameters
- frequency_penalty, include_reasoning, logit_bias, logprobs, max_tokens, min_p, presence_penalty, reasoning, reasoning_effort, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_a, top_k, top_logprobs, top_p
- Tags
- transformers, safetensors, qwen3_5, image-text-to-text
- Tool calling
- Yes
- Vision
- Yes
- Weights available
- Yes
- Weights dtype
- BF16
Release daterelease_date6
Opennessopenness4
Licenselicense1
Architecturearchitecture1
Parametersparameter_count1
Context windowcontext_length6
Max outputmax_output_tokens1
Modalitiesmodalities1
Input modalitiesmodalities_input1
Output modalitiesmodalities_output1
Tokenizertokenizer1
File sizefile_size_gb1
Hugging Face repohf_repo1
Pipeline tagpipeline_tag1
Model cardmodel_card_url1
Downloadsmetric.downloads1
Likesmetric.likes7
Aa context windowaa_context_window1
Aa opennessaa_openness1
Commercial use allowedcommercial_use_allowed1
Derivatives allowedderivatives_allowed1
Descriptiondescription1
Gatedgated1
Groq model idgroq_model_id1
Groq statusgroq_status1
Hf inference providershf_inference_providers1
Last modifiedlast_modified1
Library namelibrary_name1
Livebench hf linklivebench_hf_link1
Aa median output tokens per secondmetric.aa_median_output_tokens_per_second4
Downloads all timemetric.downloads_all_time1
Model typemodel_type1
Openrouter idopenrouter_id1
Openrouter listed atopenrouter_listed_at1
Reasoningreasoning4
Redistribution allowedredistribution_allowed1
Structured outputstructured_output1
Supported parameterssupported_parameters1
Tool callingtool_calling1
Visionvision1
Weights availableweights_available1
Weights dtypeweights_dtype1
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
Provenance
Attributed facts
45
Source tiers
T1T22 / 43
Freshest observation
2 h ago
Conflicts
None
Source documents 18
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.
Data quality (73/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →