GPQA DiamondVerified
90.5%
GPQA Diamond (AA run)
Models / Alibaba
Qwen3.8 27B
Qwen · Released 2026-08-14
Alibaba's dense open-weight Qwen3.8 sibling (14 Aug 2026). AA hosted list $0.50 / $3 per 1M tokens. Apache 2.0 weights on Hugging Face.
Self-hostable Qwen3.8 dense checkpoint for local and single-GPU runs. Hosted Qwen3.8 Max is a different product. Prefer sourced ledger rows over vendor SWE/OSWorld tables; AA scores sync from the public snapshot.
Composite66
Context262K
Input / 1M$0.50
Output / 1M$3.0
Knowledge cutoffunpublished
Output speed46 t/s
Statussolid
Family profile
Best published score in each covered benchmark family.
75/100 avg
- Reasoning
- 91
- Coding
- 47
- Agentic
- 80
- Long context
- 82
4 tested benchmark families
Editor's note
This is the dense 27B checkpoint, not hosted Qwen3.8 Max and not the 2.4T MoE. Editorial depth still needs a desk pass.Benchmark placements
Per-benchmark positions
Where Qwen3.8 27B places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.
| Benchmark | Family | Rank | Score | Source | Date |
|---|---|---|---|---|---|
| AA-LCR | Long context | #24/ 184 | 82 | Artificial Analysis | 2026-09-28 |
| Terminal-Bench | Agentic | #26/ 167 | 80 | Artificial Analysis | 2026-09-28 |
| GPQA Diamond | Reasoning | #41/ 184 | 91 | Artificial Analysis | 2026-09-28 |
| SciCode | Coding | #60/ 97 | 47 | Artificial Analysis | 2026-09-28 |
| Humanity's Last Exam | Reasoning | #66/ 186 | 34 | Artificial Analysis | 2026-09-28 |
| AA output speed | Performance | #124/ 195 | Artificial Analysis | 2026-10-12 |
Family context
Strongest rows in each family
The three highest-scoring models with pages in each capability family. Where Qwen3.8 27B shows up, it's highlighted.
ReasoningQwen3.8 27B · 91
- 196
- 295
- 395
Math
- 152
- 252
- 350
CodingQwen3.8 27B · 47
- 174
- 273
- 370
AgenticQwen3.8 27B · 80
- 191
- 289
- 389
Long contextQwen3.8 27B · 82
- 189
- 286
- 385
Newest receipts
6 recentLast 120 days
Benchmark rows added to the public ledger for Qwen3.8 27B in the last 120 days. Older rows live in the full table below.
| Benchmark | Family | Score | Source | Days ago |
|---|---|---|---|---|
| AA output speed | Performance | 46 t/s | Artificial Analysis | 10 |
| AA-LCR | Long context | 82% | Artificial Analysis | 24 |
| GPQA Diamond | Reasoning | 90.5% | Artificial Analysis | 24 |
| Humanity's Last Exam | Reasoning | 33.9% | Artificial Analysis | 24 |
| SciCode | Coding | 46.6% | Artificial Analysis | 24 |
| Terminal-Bench | Agentic | 79.78% | Artificial Analysis | 24 |
Family coverage
Best score per family
A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.
Price vs. performance
Where Qwen3.8 27B sits on the scatter
Hover any dot for name, score, and input price. Keyboard: tab through the top twelve, or use the ranking below.
AnthropicOpenAIGoogleAlibabaMetaZhipuxAIDeepSeekMiniMaxXiaomiMoonshotOtherMeituanMistralNVIDIACohere
Full benchmark ledger
Every catalog benchmark for Qwen3.8 27B. Scores link to the original source; gaps mean no public row exists yet.
6 of 13 catalog benchmarks have a sourced row for Qwen3.8 27B.
Reasoning
Humanity's Last ExamVerified
33.9%
HLE (AA run)
Coding
SciCodeVerified
46.6%
SciCode (AA run)
Agentic
Terminal-BenchVerified
79.78%
Terminal-Bench v2.1 (AA run)
Long context
AA-LCRVerified
82%
AA-LCR (AA run)
Performance
AA output speedVerified
46 t/s
Median output tokens/s (1k prompt, default provider)
Changelog
Version history
- Open weights published.
Sources
Where the numbers come from
- Hugging Face Qwen3.8-27B· checked 2026-08-24
- Alibaba Cloud Qwen3.8-27B blog· checked 2026-08-24
- Artificial Analysis Qwen3.8 27B· checked 2026-08-24
Questions this model page answers
Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.
What is Qwen3.8 27B?
Qwen3.8 27B is a Alibaba model: Alibaba's dense open-weight Qwen3.8 sibling (14 Aug 2026). AA hosted list $0.50 / $3 per 1M tokens. Apache 2.0 weights on Hugging Face.
What is Qwen3.8 27B for?
Self-hostable Qwen3.8 dense checkpoint for local and single-GPU runs. Hosted Qwen3.8 Max is a different product. Prefer sourced ledger rows over vendor SWE/OSWorld tables; AA scores sync from the public snapshot.
What does Qwen3.8 27B cost?
Qwen3.8 27B lists $0.50 input and $3.0 output per 1M tokens. Pricing last checked 2026-08-24.
How large is the Qwen3.8 27B context window?
Qwen3.8 27B publishes a 262K token context window.
What is the VerdictPal status of Qwen3.8 27B?
Qwen3.8 27B carries Solid status on the model atlas.
The Pack · Editorial newsletter
Know when the rules move.
One short email when a university changes its AI rule for theses, or a research tool changes its pricing or privacy terms. No tracking, no third-party analytics. Unsubscribe in one click.