Alibaba (Qwen) › Qwen3.8 27B › Pricing
Qwen3.8 27B pricing
API list price by Alibaba (Qwen) · checked Oct 23, 2026
$0.42 / $2.55per 1M tokens in / out
$0.085cache read · −80%
$1.13per task at medium?
#47on OpenRouter by tokens · 433.6B / week
Price sheetper 1M tokens
| Input | $0.42 | list |
| Output | $2.55 | |
| Cache read | $0.085 | −80% vs input |
| Cache write | $0.53 | +25% vs input |
| Context · max output | 262K · 131K | tokens |
Cost per task by effort?
non-thinking$2.4912.1
low$1.0514
mediumAPI default$1.1314.1
xhigh$1.0133.9
per taskHumanity's Last Exam
Same price per token at every effort; the bill grows with reasoning tokens. xhigh costs 0.9× medium for +19.8 points.
Where to buy22 providers · per 1M tokens
| Provider | Input | Output | Context |
|---|---|---|---|
| Tenstorrent | $0.02 | $0.50 | 262K |
| Darkbloom | $0.035 | $2.20 | 262K |
| Chutes | $0.039 | $1.25 | 262K |
| DekaLLM | $0.04 | $2.50 | 262K |
| Ionstream | $0.04 | $1.95 | 262K |
| Modal | $0.04 | $2.30 | 262K |
| Near AI | $0.04 | $1.35 | 262K |
| Reka | $0.04 | $2.20 | 262K |
| Wafer | $0.06 | $2.30 | 262K |
| DeepInfra | $0.20 | $2.50 | — |
| Mancer 2 | $0.20 | $2.50 | 262K |
| ModelRun | $0.20 | $2.50 | 262K |
| Phala | $0.20 | $2.50 | 1M |
| AkashML | $0.23 | $1.98 | 262K |
| Parasail | $0.24 | $2.20 | 262K |
| Featherless | $0.40 | $3 | — |
| Novita | $0.42 | $3 | — |
| Alibaba | $0.42 | $2.55 | 1M |
| Cloudflare | $0.45 | $3.20 | 262K |
| Venice | $0.45 | $3.20 | 262K |
| groq | $0.80 | $4 | — |
| Cerebras | $0.99 | $1.49 | 66K |
In Alibaba (Qwen)'s line-upper 1M in / out
| Qwen Math Plus 2024-09-19 | $0.57 / $1.72 | −10% |
| Qwen3 Next 80B A3B Instruct | $0.50 / $2 | −8% |
| Qwen3.5 Plus 2026-04-20 | $0.40 / $2.40 | −6% |
| Qwen3.8 27B | $0.42 / $2.55 | this model |
| Qwen3 VL 235B A22B | $0.53 / $2.66 | +11% |
| Qwen 3.5 122B A10B | $0.40 / $3.20 | +15% |
| Qwen 3 VL 235B A22B Thinking | $0.45 / $3.49 | +27% |
Cheaper for the same workvs Qwen 27B at medium
| GPT-6.1 Sol max | -36% | score +88% | $0.72 |
| GPT-6 Sol max | -8% | score +73% | $1.04 |
| Claude Sonnet 5.5 high | -22% | score +70% | $0.88 |
| MiMo v2.6 Pro thinkingOpen | -88% | score +68% | $0.13 |
| GPT-6 Astra low | -28% | score +66% | $0.82 |
Pricing questions
How much does Qwen3.8 27B cost?›
$0.42 per 1M input tokens and $2.55 per 1M output tokens. Cached input is $0.085. One Intelligence Index task costs about $1.13 at the default effort.
Is Qwen3.8 27B free to use?›
There is no free API tier, but the weights are open: you can run it on your own hardware for the cost of compute.
How much does prompt caching save?›
Cache reads cost $0.085 per 1M, 80% below the input price. Writing to the cache costs $0.53 per 1M.
Is Qwen3.8 27B cheaper than Qwen3.6 27B?›
Yes: $0.42 / $2.55 against $0.60 / $3.60, 29% less per token.
How does effort change the price?›
The price per token is the same at every effort. Higher effort spends more reasoning tokens: a task costs $2.49 at non-thinking, $1.05 at low, $1.13 at medium, $1.01 at xhigh.
What is a cheaper alternative to Qwen3.8 27B?›
GPT-6.1 Sol at max effort costs -36% per task with a score +88% on Humanity's Last Exam.