← Model list

Qwen3.8 27B

alibaba·alibaba/qwen3.8-27b·BETA·Open weights·qwen series·NEW

Dense 27B vision-language model for coding, agent tasks, and image and video understanding

At a glance
Reference price$0.575 / $3.45 per 1M
Cache read $0.115 · Kilo Gateway
Lowest paid$0.10 / $0.40
RunInfra Gateway · 5.7× spread
1 more $0 channels
Context262,144
⚠ Providers report 262,144–1,000,000; the table below is authoritative
Output limit131,072
Capabilities
ReasoningTool useStructured outputTemperatureAttachments
⚠ Providers report capability flags inconsistently
Modalities
TextImageVideoPDF
Knowledge cutoff
Released / updated2026-08-14 / 2026-08-14

Quality & performanceArtificial Analysis · Intelligence Index v4.1

Intelligence52
Value297
Coding68.1
Agentic50.9
Output speed
TTFT
Value formulaIQ 52 ÷ min blended $0.175 = 297

Quality is independently evaluated by Artificial Analysis. Speed/latency are model-level medians.

Available at 14 providers13 with public prices · 1 free

ProviderTierInputOutputCache readCache writeContextOutput limitStatus
Hetzner
Qwen3.8-27B
CloudFree$0262,144262,144beta
RunInfra
Qwen/Qwen3.8-27B
Gateway$0.10$0.40$0.01262,14432,768
EmpirioLabs AI
qwen3-8-27b
Gateway$0.17$0.50$0.08262,14432,768
NanoGPTGateway$0.20$1.40$0.04262,14432,768
Kosmik ComputeGateway$0.35$2.20$0.09262,14432,768
CortecsGateway$0.334$2.45$0.111262,144262,144
Deep Infra
Qwen/Qwen3.8-27B
Cloud$0.40$3.00$0.04262,14432,768
Weights & Biases
Qwen/Qwen3.8-27B
Cloud$0.40$3.00$0.15262,144262,144
Eden AIGateway$0.50$3.00$0.10$0.6251,000,00032,768
Cloudflare Workers AICloud$0.45$3.20$0.05262,144262,144

Sorted by blended price (input×0.75 + output×0.25) asc. The official channel always shows regardless of rank. Whether a gateway's low price is actually usable can't be verified.

Your usage cost

1RunInfra$29.20
2EmpirioLabs AI$48.20
3NanoGPT$90.80
4Kosmik Compute$148.80
5Cortecs$162.59
6Deep Infra$186.80

1 more channels offer $0 (Hetzner); free tiers usually have rate limits and no SLA, excluded from ranking.

The cheapest paid channel is the only channel.

Benchmark1 items

NameConditionsScoreMetricSource
SWE-bench Pro61.7resolvedSource ↗

The same benchmark scores very differently across harness / dataset, so the qualifying conditions must be shown together.

Reasoning control

effort = noneeffort = loweffort = mediumeffort = xhighToggle (on / off)

2 / 14 providers expose no reasoning control (reasoning_options: []).

Related models

Price historyone sample accumulated per data sync

Input listOutput listMin blended
$3.45$02026-08-152026-08-212026-08-15 · Input list $0.452026-08-17 · Input list $0.452026-08-20 · Input list $0.552026-08-21 · Input list $0.5752026-08-15 · Output list $3.202026-08-17 · Output list $3.202026-08-20 · Output list $3.302026-08-21 · Output list $3.452026-08-15 · Min blended $1.142026-08-17 · Min blended $0.1752026-08-20 · Min blended $0.1752026-08-21 · Min blended $0.175