Qwen3.8 27B

AlibabaMCP Tool CallReasoning
alibaba/qwen3.8-27bOpen Weights

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

Benchmark Pricing$0.10 / 1M in$0.40 / 1M out$0.04 cache read
Context: 262,144 tokens•Max Output: 32,768 tokens•Released: 2026-08-14

TECHNICAL SPECIFICATIONS

Context Window262,144262.1K tokens
Max Output32,76832.8K tokens
LicenseOpen SourceSelf-hostable
Input Modalities
TextVisionVideo
Function / Tool CallingSupported (MCP Ready)
Structured OutputsGuaranteed Schema

MODEL CONTEXT PROTOCOL (MCP) READINESS

✓ Qwen3.8 27B natively supports OpenAI / Anthropic format tool calling and is fully compatible with MCP servers and agent loops.

MCP Context Window Recommendations

With a context capacity of 262,144 tokens, this model can handle moderately large MCP tool definitions and multi-turn conversation logs comfortably.

STANDARDIZED BENCHMARKS & EVALS

SWE-bench Proresolved
61.7
Terminal-Benchscore
73
NL2Reposcore
42.3
DeepSWEscore
42.2
QwenSWEBenchavg@3
79
CoWorkBenchscore
70.7
JobBenchscore
33.4
Agents' Last Exampass@1
20.4
Agents' Last Examscore
42.9
IFBenchscore
79.5
GPQA Diamondscore
89.2
Humanity's Last Examscore
30.8
LiveCodeBenchscore
90.3
OSWorld-Verifiedscore
84.3
WebArena-Verifiedscore
64.8
AndroidWorldscore
81.9
SWE-MMscore
38.6
RealWorldQAscore
85.9
ERQAscore
65.5
RecreationBenchscore
47.1
ClawEval-MMpass@3
57.4
ClawEval-MMaverage score
56.9
Vision2Webscore
62.9
MathVisionscore
90
MathVisionscore
94.6
BabyVisionscore
65.7
BabyVisionscore
85.6
CharXiv Reasoningscore
83.7
CharXiv Reasoningscore
90.2
OmniDocBenchscore
91.1

API CALL INVOCATION EXAMPLE

// Example: Calling Qwen3.8 27B via OpenAI SDK
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ALIBABA_API_KEY,
});

const response = await client.chat.completions.create({
  model: "alibaba/qwen3.8-27b",
  messages: [{ role: "user", content: "Hello! Can you inspect MCP tools?" }],
});

console.log(response.choices[0].message.content);

PRICING BY PROVIDER

15 providers
Cortecsqwen3.8-27b
$0.10 / $0.40$0.04 cache
AMDQwen3.8-27B
$0.00 / $0.00$0.00 cache
NanoGPTqwen/qwen3.8-27b-novelist
$0.10 / $0.60$0.05 cache
RunInfraQwen/Qwen3.8-27B
$0.10 / $0.40$0.01 cache
Scalewayqwen3.8-27b
$0.68 / $3.76$0.14 cache
Vultrqwen3.8-27b
$0.15 / $1.00
Charm Hyperqwen3.8-27b
$0.50 / $3.00$0.10 cache
Novita AIqwen/qwen3.8-27b
$0.42 / $3.00$0.09 cache
ai&qwen/qwen3.8-27b
$0.40 / $3.00$0.20 cache
Cerebrasqwen-3.8-27b
$0.99 / $1.49$0.99 cache
LLM Gatewaynovita/qwen3.8-27b
$0.42 / $3.00$0.09 cache
IteraComputeqwen/qwen3.8-27b
$0.00 / $0.00$0.00 cache
CoreWeaveQwen/Qwen3.8-27B
$0.40 / $3.00$0.15 cache
Tempr Gatewaycerebras/qwen-3.8-27b
$0.99 / $1.49
AI/ML APIalibaba/qwen3.8-27b
$0.62 / $4.40

Rates in USD per 1M tokens. Rates synchronized from models.dev public catalog.

Prompt Caching Active

This model supports prompt caching at $0.04 / 1M tokens, saving up to 60% on repetitive MCP tool definitions and system instructions.