Effort: xhigh · Xhigh
QwenReleased Aug 14, 2026
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 enabled or disabled.
Context window1M
Max output235.9K
Parameters27B
Provider endpoints28
Benchmark results19
Measured performance
All leaderboards ↗Benchmark results
Effort: xhigh · Xhigh
GDPval-AA ↗2.1
1,419.88Effort: xhigh · Xhigh
AutomationBench-AA ↗AA Index 4.3.2 collection
48.24%Effort: xhigh · Xhigh
Effort: xhigh · Xhigh · Harness unreported
SciCode ↗AA Index 4.3.2 collection
46.64%AA · Under reviewEffort: xhigh · Xhigh
Humanity’s Last Exam ↗AA Index 4.3.2 collection
33.92%Effort: xhigh · Xhigh
GDP.pdf ↗AA Index 4.3.2 collection
16.4%Effort: xhigh · Xhigh
CritPt ↗AA Index 4.3.2 collection
5.43%AA · Under reviewEffort: xhigh · Xhigh
AA-Omniscience ↗AA Index 4.3.2 collection
-9.98Effort: xhigh · Xhigh
AA-LCR ↗1.1
82%Effort: xhigh · Xhigh
LiveBench Overall ↗2.0 · 2026-10-14 (dataset 2026-09-02)
62.66%Effort: max · single run · equal mean of 7 suites
LiveBench Design ↗2.0 · 2026-10-14 (dataset 2026-09-02)
75.9%Effort: max · single run
LiveBench Research ↗2.0 · 2026-10-14 (dataset 2026-09-02)
59.4%Effort: max · single run
LiveBench Browser Use ↗2.0 · 2026-10-14 (dataset 2026-09-02)
32%Effort: max · single run
LiveBench Coding ↗2.0 · 2026-10-14 (dataset 2026-09-02)
86.8%Effort: max · single run
LiveBench Agentic ↗2.0 · 2026-10-14 (dataset 2026-09-02)
27.9%Effort: max · single run
LiveBench Data Analysis ↗2.0 · 2026-10-14 (dataset 2026-09-02)
76.6%Effort: max · single run
LiveBench Reasoning ↗2.0 · 2026-10-14 (dataset 2026-09-02)
80%Effort: max · single run
Provider measurements
1 configurationSpeed
≈10K input tokens1 requestmedian / 72h
Artificial Analysis · first-party providers · serving endpoint not reported
Alibaba
Qwen3.8 27B (Xhigh)
Reasoning xhigh
- Output rate
- 41.3222 tok/s
- Time to first chunk
- 3.98625 s
Last checked Source ↗
Methodology 2.2.0 & measurement details
First chunk includes request processing and network time, and may include reasoning.
- Prompt
- long · ≈10,000 input tokens
- Parallel requests
- 1
- Minimum answer
- 1,500 tokens
- Aggregation
- median over 72 hours
- Token counting
- o200k_base
- Benchmark client
- Google Cloud us-central1-a
- Serving endpoint
- Not reported
- Serving region
- Not reported
- Sample size
- Not reported
- Window end
- Not reported
Access & cost
28 quotesPrices across providers
Your token mix
USD · listed token ratesOpenRouter
Catalog quote- Input
- $0.24
- Output
- $2.20
- Cached input
- $0.05
novita
via Hugging Face- Input
- $0.42
- Output
- $3.00
cerebras
via Hugging Face- Input
- $0.99
- Output
- $1.49
featherless-ai
via Hugging Face- Input
- —
- Output
- —
ovhcloud
via Hugging Face- Input
- $0.47
- Output
- $3.19
deepinfra
via Hugging Face- Input
- $0.20
- Output
- $2.50
- Input
- $0.20
- Output
- $2.50
Groq
Preview · evaluation only · Developer PAYG
- Input
- $0.80
- Output
- $4.00
Preview models may be discontinued at short notice.
- Input
- $0.02
- Output
- $0.50
- Cached input
- $0.009
- Input
- $0.035
- Output
- $2.20
- Cached input
- $0.0175
- Input
- $0.039
- Output
- $1.25
- Cached input
- $0.0039
- Input
- $0.04
- Output
- $1.35
- Cached input
- $0.018
- Input
- $0.04
- Output
- $2.20
- Cached input
- $0.02
- Input
- $0.04
- Output
- $2.30
- Cached input
- $0.039
- Input
- $0.04
- Output
- $2.50
- Cached input
- $0.02
- Input
- $0.0893
- Output
- $2.35
- Cached input
- $0.085
- Input
- $0.15
- Output
- $1.875
- Cached input
- $0.0375
Source discount 25%; rates as published
- Input
- $0.20
- Output
- $2.50
- Cached input
- $0.09
- Input
- $0.20
- Output
- $2.50
- Input
- $0.20
- Output
- $2.50
- Cached input
- $0.05
- Input
- $0.225
- Output
- $1.98
- Cached input
- $0.05
- Input
- $0.24
- Output
- $2.20
- Cached input
- $0.05
- Input
- $0.40
- Output
- $3.00
- Cached input
- $0.15
Endpoint degraded (status -5)
- Input
- $0.42
- Output
- $3.00
- Cached input
- $0.085
- Input
- $0.425
- Output
- $2.55
- Cached input
- $0.085
Cache write $0.53125/M tokens
- Input
- $0.45
- Output
- $3.20
- Cached input
- $0.05
- Input
- $0.45
- Output
- $3.20
- Input
- $0.99
- Output
- $1.49
- Cached input
- $0.99
Model specifications
- Developer
- Qwen
- Modality
- Language
- Context
- 1,000,000 tokens
- Maximum output
- 235,929 tokens
- Weights
- Open weights
- Released
- Aug 14, 2026
- Provider listed
- Aug 5, 2026
- First discovered
- Oct 16, 2026
Exact identities
- openrouter
qwen/qwen3.8-27b - huggingface
Qwen/Qwen3.8-27B - deepinfra
Qwen/Qwen3.8-27B - artificial-analysis
b01dee41-c62b-48ed-8d16-984adc405e5c - livebench
qwen3.8-27b
Sources
Updated Oct 17, 2026
More from Qwen
View all ↗QwenListed Sep 14
Qwen3.8-Flash
Qwen's fast, low-cost model with a one-million-token context window, billed at a flat rate across the entire window.
1M contextNo reported score
No reported score
QwenListed Aug 3
Qwen3.8-Max
Qwen's latest 2.4-trillion-parameter MoE flagship model delivering a comprehensive leap in coding and professional work.
256K contextNo reported score
No reported score
QwenListed May 28
Qwen3.7-Max
The largest and most capable in the Qwen3.7 series. Qwen3.7 is a next‑generation flagship model designed for the agent‑centric.
256K contextNo reported score
No reported score
QwenListed Apr 15
Qwen3.6-35B-A3B
Qwen3.6-35B-A3B is Alibaba's latest flagship Mixture-of-Experts model, with 35B total parameters and only 3B activated per token (256 experts, 8 routed + 1 shared). Built on direct feedback from the…
262.1K context35BImage inputVideo inputNo reported score
No reported score
$0.05in$0.70out/ 1M tokens · USDDarkbloom · via OpenRouter · 15 endpointsdarkbloom/fp4 · quantization=fp4