Alibaba Open weights: Qwen/Qwen3.8-27BReasoning modelAuto-detected

Qwen 3.8 27B

Name: Qwen 3.8 27B benchmark results
Creator: BenchLeader

Best configuration ranks #148 of 610 on the BenchLeader Index at 57.1 ±4.9. Last measured 8 Sept 2026. Released 14 Aug 2026.

Blended price: $1.13/M; $0.500 in · $3.00 out
Output speed: 46 tok/s
First answer: 47 s; first token 1.48 s
Context: 262k

How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.

Overall index57
Reasoning63
Coding66
Agents & tools48
Maths40
Knowledge59
Human preference63
Multimodal62
Long context68
Composite62

Reasoning-effort configurations

The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.

Effort	Index	Rank	Speed	First answer	Chat reply cost	Categories
no reasoning	52.6	#245	54 tok/s	3.87 s	$0.0011	Coding 38 · Composite 57 · Knowledge 60 · Long context 61 · Multimodal 54
low	54.3	#203	52 tok/s	42 s	$0.0011	Coding 44 · Composite 66 · Knowledge 51 · Long context 65 · Multimodal 58
medium	53.9	#214	50 tok/s	44 s	$0.0011	Coding 41 · Composite 68 · Knowledge 47 · Long context 66 · Multimodal 59
xhigh	52.0	#260	46 tok/s	47 s	$0.0015	Agents & tools 47 · Coding 57 · Knowledge 55
defaultbest	57.1	#148	46 tok/s	47 s	$0.0011	Agents & tools 48 · Coding 66 · Composite 62 · Human preference 63 · Knowledge 59 · Long context 68 · Maths 40 · Multimodal 62 · Reasoning 63

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Reasoning

Benchmark	no reasoning	low	medium	xhigh	default	Source
LMArena Hard Prompts	–	–	–	–	1460#79	LMArena
LiveBench Reasoningnot in index	–	–	–	–	80.0%#39	LiveBench
GPQA Diamond (AA)not in index	81.8%#172	84.5%#136	84.5%#136	–	90.5%#59	Artificial Analysis
Humanity's Last Exam (AA)not in index	12.1%#218	14.0%#201	14.1%#199	–	33.9%#85	Artificial Analysis
GPQA Diamond (Vals)not in index	–	–	–	88.9%#31	–	Vals AI

Coding

Benchmark	no reasoning	low	medium	xhigh	default	Source
SciCode	35.6%#131	39.8%#111	38.1%#121	44.7%#87	–	SciCode
LMArena Coding	–	–	–	–	1494#72	LMArena
LMArena WebDev	–	–	–	–	1592#15	LMArena
LiveBench Codingnot in index	–	–	–	–	75.7%#36	LiveBench
SciCode (AA)not in index	36.2%#133	40.0%#116	39.0%#119	–	46.6%#85	Artificial Analysis
LiveCodeBench	–	–	–	84.0%#41	–	Vals AI
SWE-bench (Vals)not in index	–	–	–	86.0%#15	–	Vals AI

Agents & tools

Benchmark	no reasoning	low	medium	xhigh	default	Source
LMArena Agent	–	–	–	–	0.1#25	LMArena
LiveBench Agentic Codingnot in index	–	–	–	–	61.4%#9	LiveBench
Terminal-Bench 2.1 (Vals)	–	–	–	58.4%#32	–	Vals AI

Maths

Benchmark	no reasoning	low	medium	xhigh	default	Source	Trend
ProofBench	–	–	–	–	16.0%#41	Vals AI
LiveBench Mathematicsnot in index	–	–	–	–	86.2%#38	LiveBench

Knowledge

Benchmark	no reasoning	low	medium	xhigh	default	Source
LiveBench Data Analysisnot in index	–	–	–	–	76.6%#22	LiveBench
AA-Omniscience	-8.0#127	-26.7#204	-36.1#239	–	-10.0#139	Artificial Analysis
MMLU-Pro	–	–	–	84.3%#62	–	Vals AI
LegalBench	–	–	–	82.4%#64	–	Vals AI
TaxEval	–	–	–	70.8%#83	–	Vals AI

Instruction following

Benchmark	no reasoning	low	medium	xhigh	default	Source	Trend
LiveBench Languagenot in index	–	–	–	–	74.3%#41	LiveBench

Human preference

Benchmark	no reasoning	low	medium	xhigh	default	Source	Trend
LMArena Text	–	–	–	–	1436#84	LMArena

Multimodal

Benchmark	no reasoning	low	medium	xhigh	default	Source	Trend
LMArena Vision	–	–	–	–	1279#28	LMArena
MMMU-Pro	69.9%#119	73.8%#93	74.2%#87	–	76.3%#64	Artificial Analysis

Long context

Benchmark	no reasoning	low	medium	xhigh	default	Source	Trend
AA-LCR	69.3%#170	77.3%#91	79.7%#59	–	82.0%#25	Artificial Analysis

Composite

Benchmark	no reasoning	low	medium	xhigh	default	Source
LiveBench	–	–	–	–	75.3%#28	LiveBench
AA Intelligence Index	22.4#146	29.2#84	30.7#76	–	33.9#62	Artificial Analysis
Vals Indexnot in index	–	–	–	48.5#29	–	Vals AI

Where to run it

Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.

Provider	Speed	First token	Input $/M	Output $/M	Context	Quantisation
Venice	87 tok/s	1.49 s	$0.450	$3.20	262k	fp8
AkashML	64 tok/s	0.74 s	$0.250	$2.20	262k	fp8
Phala	53 tok/s	1.08 s	$0.300	$3.00	262k	–
CoreWeave	53 tok/s	0.42 s	$0.400	$3.00	262k	fp8
Ionstream	51 tok/s	0.73 s	$0.350	$2.55	262k	fp8
Reka AI	43 tok/s	0.69 s	$0.214	$2.55	262k	fp8
Parasail	43 tok/s	1.48 s	$0.240	$2.20	262k	fp8
Alibaba Cloud Int.	43 tok/s	1.45 s	$0.425	$2.55	1M	–
Chutes	30 tok/s	2.17 s	$0.320	$2.50	262k	fp8
NovitaAI	25 tok/s	1.53 s	$0.420	$3.00	1M	–
Cloudflare	25 tok/s	1.62 s	$0.450	$3.20	262k	–
io.net	22 tok/s	2.42 s	$0.300	$2.80	66k	fp8
Darkbloom	16 tok/s	1.74 s	$0.150	$2.00	262k	fp4

Price history

Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.

input outputnow $0.350 in · $2.55 out

What a task costs

Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache at $0.040 per 1M. Reasoning tokens are not modelled.

Workload	Tokens in / out	Cost	With caching	Time
Chat reply	400 / 300	$0.0011	$0.0010	53.9 s
Summarise a 30-page report	12,000 / 600	$0.0078	$0.0037	1.0 min
Code edit	6,000 / 1,500	$0.0075	$0.0054	1.3 min
Agentic coding session	60,000 / 4,000	$0.042	$0.021	2.2 min
Structured extraction	2,000 / 200	$0.0016	$0.0009	51.8 s