// npm install @heyputer/puter.js
import { puter } from '@heyputer/puter.js';
puter.ai.chat("Explain quantum computing in simple terms", {
model: "qwen/qwen3.5-4b"
}).then(response => {
document.body.innerHTML = response.message.content;
});
<html>
<body>
<script src="https://js.puter.com/v2/"></script>
<script>
puter.ai.chat("Explain quantum computing in simple terms", {
model: "qwen/qwen3.5-4b"
}).then(response => {
document.body.innerHTML = response.message.content;
});
</script>
</body>
</html>
# pip install openai
from openai import OpenAI
client = OpenAI(
base_url="https://api.puter.com/puterai/openai/v1/",
api_key="YOUR_PUTER_AUTH_TOKEN",
)
response = client.chat.completions.create(
model="qwen/qwen3.5-4b",
messages=[
{"role": "user", "content": "Explain quantum computing in simple terms"}
],
)
print(response.choices[0].message.content)
curl https://api.puter.com/puterai/openai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_PUTER_AUTH_TOKEN" \
-d '{
"model": "qwen/qwen3.5-4b",
"messages": [
{"role": "user", "content": "Explain quantum computing in simple terms"}
]
}'
Model Card
Qwen3.5-4B is a compact multimodal model from Alibaba's Qwen3.5 Small series, released alongside 0.8B, 2B, and 9B siblings in early 2026 as part of the family succeeding Qwen3.
Unlike models that bolt a vision tower onto a text backbone, it processes text and images in a unified latent space, using a hybrid architecture that interleaves linear-attention (Gated DeltaNet) blocks with standard attention. Alibaba reports 79.1% on MMLU-Pro, 76.2% on GPQA Diamond, and 74-77% on HMMT, scores that Qwen says close the gap with much larger models.
Qwen positions the 4B size as a multimodal base for lightweight agents rather than a pure edge model like its 0.8B and 2B siblings. Via this API it offers a 256K context window and 32K max output tokens, a fit for developers who want multimodal reasoning and OCR at low per-token cost.
Context Window 256K
tokens
Max Output 33K
tokens
Input Cost $0.04
per million tokens
Output Cost $0.07
per million tokens
Release Date N/A
Output Speed 17
tokens / sec
Latency 0.48s
time to first token
Model Playground
Try Qwen3.5-4B instantly in your browser.
This playground uses the Puter.js AI API — no API keys or setup required.
Benchmarks
How Qwen3.5-4B performs on standard evaluations.
| Benchmark | Score |
|---|---|
| GPQA Diamond Graduate-level science Q&A | 77.1% |
| Humanity's Last Exam Cross-domain reasoning | 9.9% |
| IFBench Instruction following | 52.0% |
| LCR Long-context reasoning | 63.0% |
| Terminal-Bench Hard Agentic terminal tasks | 18.2% |
| τ²-Bench Tool use / agents | 92.1% |
Scores sourced from Artificial Analysis.
Find other Qwen models →
Qwen3.8 Max (0902)
Qwen3.8 Max (0902) is a September 2026 snapshot of Alibaba's Qwen3.8 Max, a 2.4-trillion-parameter mixture-of-experts flagship with roughly 95 billion active parameters per token. It keeps the 1M-token context window, thinking mode, and tool ecosystem of the base model, with additional post-training on coding and office-work tasks. On Code Arena WebDev it ranks first overall at 1,691 points, ahead of Claude Opus 5 Max (1,687) and Kimi K3 Max (1,674). The largest gains are on TerminalBench 3.0 and ProgramBench Almost Solved, both more than doubling versus the prior snapshot. Reported core scores include GPQA Diamond 92.6 and PaperBench 93.0. Against Claude Opus 5, it leads on several coding benchmarks and on WorkArena, though Opus 5 still leads on most agentic coding rows and both office-work benchmarks. Pricing is unchanged from the base Qwen3.8 Max, making it a straightforward upgrade for coding and office-automation workloads already on the platform.
ChatQwen3.8 Flash
Qwen3.8 Flash is a multimodal model from Alibaba's Qwen team, released August 26, 2026, as the fast, lower-cost tier of the Qwen3.8 family alongside Qwen3.8 Max and Qwen3.8 27B. It uses a mixture-of-experts architecture with 125B total parameters and 6B active per token, an early preview of the architecture planned for Qwen4. It accepts text, image, and video input and returns text, with a 1,000,000 token context window and output capped at 128,000 tokens. The API supports tool calling, structured outputs via JSON schema, and prompt caching, with cached input billed at $0.016 per million tokens. Pricing is $0.06 per million input tokens and $0.19 per million output tokens, about one-twelfth the cost of Qwen3.8 Max. Alibaba says it was trained at roughly one-ninth the cost of Qwen3.7-Plus and reports higher scores on benchmarks including SWE-bench Pro and CoWorkBench, an agentic office-task benchmark.
ChatQwen3.8 27B
Qwen3.8 27B is a dense, open-weight multimodal model from Alibaba's Qwen team, released August 14, 2026 as a smaller member of the Qwen3.8 family alongside the flagship Qwen3.8 Max. It combines Gated DeltaNet linear attention with standard gated attention across 64 layers, giving a 27 billion parameter dense model a native 262K token context window, extendable to 1M tokens. It accepts text, image, and video input, including hour-scale video and STEM diagrams. Alibaba reports 61.7 on SWE-bench Pro and 73.0 on Terminal Bench 2.1, both improvements over the earlier Qwen3.6 27B, and 89.2 on GPQA Diamond. Released under Apache 2.0, it gives developers an open-weight alternative to Qwen3.8 Max for coding and agentic tasks, at a fraction of the parameter count.
Frequently Asked Questions
You can access Qwen3.5-4B by Qwen through Puter.js AI API. Include the library in your web app or Node.js project and start making calls with just a few lines of JavaScript — no backend and no configuration required. You can also use it with Python or cURL via Puter's OpenAI-compatible API.
Yes, it is free if you're using it through Puter.js. With the User-Pays Model, you can add Qwen3.5-4B to your app at no cost — your users pay for their own AI usage directly, making it completely free for you as a developer.
| Price per 1M tokens | |
|---|---|
| Input | $0.04 |
| Output | $0.07 |
Qwen3.5-4B supports a context window of 256K tokens. For reference, that is roughly equivalent to 512 pages of text.
Qwen3.5-4B can generate up to 33K tokens in a single response.
Qwen3.5-4B scores 13.9 on the Artificial Analysis Intelligence Index, outperforming 55% of tracked models. On coding, it scores 22.6 (outperforms 25% of models).
Yes — the Qwen3.5-4B API works with any JavaScript framework, Node.js, or plain HTML through Puter.js. Just include the library and start building. See the documentation for more details.
Get started with Puter.js
Add Qwen3.5-4B to your app without worrying about API keys or setup.
Read the Docs View Tutorials