Qwen3 VL 8B Thinking

qwen/qwen3-vl-8b-thinking
VisionToolsJSONReasoning
by Qwen · 2025-10-14

Qwen3-VL 8B Thinking — open-weight small vision-language reasoning model, 8B params, 128k context.

ctx131.1K tokens
Max output41K
Inputtext + image + video
Outputtext
p50 TTFT4.25 s
INPUT$0.18/ 1M tokens
OUTPUT$2.10/ 1M tokens
p50 TTFT4.25 s7d
p95 TTFT8.55 s7d
TRAFFIC146.2Ktokens / 7d

Qwen3 VL 8B Thinking is an 8-billion-parameter multimodal language model developed by the Qwen team at Alibaba Cloud, hosted on OrcaRouter under the provider qwen. It belongs to the Qwen3 family of…

What is Qwen3 VL 8B Thinking?

Who should use this model?

What modalities does it support?

How does the 'Thinking' variant differ from standard Qwen3 VL?

Code samples

Call from any SDK

OpenAI-compatible — keep the SDK you already use

  • OpenAI SDKhttps://api.orcarouter.ai/v1
from openai import OpenAI

client = OpenAI(
    base_url="https://api.orcarouter.ai/v1",
    api_key="$ORCAROUTER_API_KEY",
)

response = client.chat.completions.create(
    model="qwen/qwen3-vl-8b-thinking",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Supported parameters

  • enable_search
  • enable_thinking
  • include_reasoning
  • logprobs
  • max_tokens
  • n
  • parallel_tool_calls
  • presence_penalty
  • reasoning
  • repetition_penalty
  • response_format
  • seed
  • stop
  • stream
  • stream_options
  • temperature
  • thinking_budget
  • tool_choice
  • tools
  • top_k
  • top_logprobs
  • top_p

Pricing

Input / 1M tokens$0.180
Output / 1M tokens$2.10
CurrencyUSD

Cost calculator

Tokens / month10MM
Input share70%%
Estimated / month $7.56

Estimate based on list price

Token & cost estimator

Input tokens: 20Cost per request: $0.001054

Estimate only — actual token counts depend on the provider's tokenizer.

Performance

p50 TTFT
4.25 s
Output speed
71.2 tok/s
p95 TTFT
8.55 s
Error rate
0%

Public benchmarks

Source: Design Arena

How it compares

Qwen3 VL 8B Thinkingqwen/qwen3-max-previewQwen3.5 397B A17Bqwen/qwen3.5-plus
Input $/M$0.18$0.86$0.17$0.12
Output $/M$2.10$3.44$1.03$0.69
Context131K262K33K1.0M
Quality4/108/108/108/10
Compare side-by-sideCompare side-by-sideCompare side-by-sideCompare side-by-side

FAQ

What is the cost per million tokens for Qwen3 VL 8B Thinking on OrcaRouter?
Input tokens: $0.18 per 1 million tokens. Output tokens: $2.10 per 1 million tokens. These are the provider's rates passed through with zero markup.
What is the context window and maximum output token count?
Context window is 131,072 tokens. Maximum output tokens is 40,960 tokens.
What are the main strengths of this model?
It handles three input modalities (text, image, video), offers a large context and output length, and includes a thinking (chain-of-thought) capability that improves performance on complex reasoning tasks.
How does this model compare to Qwen2 VL 7B?
It has a larger context (131K vs 32K), larger max output (40K vs 2K), and adds thinking capabilities. Pricing is slightly higher but offers improved reasoning and multimodal understanding.
Does OrcaRouter cache any user data when using this model?
OrcaRouter does not report any caching mechanism that stores prompts or images; data handling follows standard API practices. Consult OrcaRouter's privacy policy for details.
How do I call this model via an OpenAI-compatible API?
Send a POST to https://api.orcarouter.ai/v1/chat/completions with model "qwen/qwen3-vl-8b-thinking" and your API key. Use a content array with text and image_url entries for multimodal input.
Can the model process video input?
Yes, video is accepted by sending extracted frames as separate image_url entries. The model has no native video codec support; it works on static frame sets.
Is there any restriction on the number of images per request?
No, except the total token count (including text and image tokens) must be within the 131,072 context limit. There is no separate image limit.
What programming languages or libraries can I use to access this model via OrcaRouter?
Any language that supports HTTP requests and the OpenAI API format. Python with the openai library, JavaScript with fetch, etc. Just set the base URL and model ID.
Does the thinking variant always return chain-of-thought reasoning?
The model internally uses step-by-step reasoning before generating the final answer, but the output you see is the complete response. You cannot extract the intermediate thinking tokens separately.

Embed this badge

Qwen: Qwen3 VL 8B Thinking$0.18/M in4250ms p50via OrcaRouter
HTML <a href="https://www.orcarouter.ai/models/qwen/qwen3-vl-8b-thinking" target="_blank"> <img src="https://www.orcarouter.ai/embed/qwen/qwen3-vl-8b-thinking.svg" alt="Qwen: Qwen3 VL 8B Thinking on OrcaRouter" /> </a>
Markdown [![Qwen: Qwen3 VL 8B Thinking](https://www.orcarouter.ai/embed/qwen/qwen3-vl-8b-thinking.svg)](https://www.orcarouter.ai/models/qwen/qwen3-vl-8b-thinking)

Model card as data

GET /api/public/models/qwen/qwen3-vl-8b-thinkingOpen
Machine-readable:/llms.txt/llms-full.txt