impulse

Qwen3 VL 8B Instruct

Open weights

Qwen3-VL-8B-Instruct is a multimodal vision-language model from the Qwen3-VL series, built for high-fidelity understanding and reasoning across text, images, and video.

API id qwen3-vl-8b-instructReleased 14 Oct 2025 (Oct 2025)
Input / 1M
$0.08
at Novita AI
Output / 1M
$0.50
at Novita AI
Context window
131K
tokens
Max output
33K
tokens per reply
Hosts
3
3 listings
Released
Oct 2025
14 Oct 2025

Pricing

List prices at Novita AI, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.08$0.50

What it costs a month

At Novita AI list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out$61.50
  • RAG over documents2,000 requests a day, 8K in / 0.6K out$56.40
  • Coding agent1,000 requests a day, 40K in / 2K out$126
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$312
Use your own numbers

Where to run it

3 hosts serve Qwen3 VL 8B Instruct. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
Novita AIInference
qwen3-vl-8b-instruct
$0.08 / $0.50
Fireworks AIInference
qwen3-vl-8b-instruct
$0.20 / $0.20
+8% vs headline
Together AIInference
Qwen3-VL-8B-Instruct
$0.18 / $0.68
+65% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
TextImages
Produces
Text
Features
Parallel tool calls
Open weights
License: Apache-2.0
Runs in about 21 GB of GPU memory
Runtimes: Ollama, SGLang, llama.cpp, vLLM

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

No host sells it for less than Novita AI.

Smaller Alibaba Qwen models

No cheaper Alibaba Qwen model keeps its key capabilities.

Other frontier labs

No cheaper match from another frontier lab.

Open-weight picks

  • Qwen3.8 Flash
    Alibaba Qwen · on AIHubMix
    $0.113 / $0.38
    Has every key capability
    3% less
  • Nemotron 3.5 Super VL
    NVIDIA · on Kyma API
    $0.081 / $0.472
    Has every key capability
    3% less
  • Gemma 4 31B IT
    Google · on W&B Inference
    $0.10 / $0.34
    Has every key capability
    14% less
  • Ministral 8B 2512
    Mistral AI · on Mistral AI
    $0.15 / $0.15
    Has every key capability
    19% less

Call it

Using Novita AI's API.

Python
from litellm import completion

response = completion(
    model="novita/qwen3-vl-8b-instruct",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does Qwen3 VL 8B Instruct cost?

On Novita AI, Qwen3 VL 8B Instruct costs $0.08 per million input tokens and $0.50 per million output tokens.

Which APIs offer Qwen3 VL 8B Instruct?

Qwen3 VL 8B Instruct runs on 3 APIs: Novita AI, Fireworks AI and Together AI.

What is Qwen3 VL 8B Instruct's context window?

131K tokens, with up to 33K tokens in one reply.

Is Qwen3 VL 8B Instruct being retired?

No retirement date is scheduled for it on any API we track.

What are cheaper alternatives to Qwen3 VL 8B Instruct?

Models that cost less and keep most of what it can do: Qwen3.8 Flash on AIHubMix (3% less). Savings are on a mix of 3 input tokens to 1 output token.

When was Qwen3 VL 8B Instruct released?

Alibaba Qwen released it on 14 Oct 2025.

What is Qwen3 VL 8B Instruct's API model id?

Use qwen3-vl-8b-instruct with Novita AI. Other hosts use their own ids: Qwen3-VL-8B-Instruct on Together AI.

Sources

Raw catalog entry for qwen3-vl-8b-instruct on Novita AI
{
  "litellm_provider": "novita",
  "mode": "chat",
  "input_cost_per_token": 8e-8,
  "output_cost_per_token": 5e-7,
  "max_input_tokens": 131072,
  "max_output_tokens": 32768,
  "supports_function_calling": true,
  "supports_parallel_function_calling": true,
  "supports_tool_choice": true,
  "supports_vision": true,
  "supports_system_messages": true,
  "supports_response_schema": true
}