ColinBuilds.com · Alibaba Cloud Model Studio hosted pricing
Qwen3-Coder-480B-A35B-Instruct
- Input / 1M
- $1.50
- Output / 1M
- $7.50
- Native 256K context (1M with extrapolation)
- 262,144 tokens
Current Alibaba Cloud Model Studio international pricing for qwen3-coder-480b-a35b-instruct; 0-32K token tier used for calculator; larger context tiers cost more · as of June 30, 2026
Compare models
Same per-1M input/output rates as above. Enter your monthly token volumes to compare bills.
Qwen3-Coder-480B-A35B-Instruct
- Input: 1M × $1.50 = $1.50
- Output: 1M × $7.50 = $7.50
- Total
- $9.00
GPT-4o
- Input: 1M × $2.50 = $2.50
- Output: 1M × $10.00 = $10.00
- Total
- $12.50
Difference
$3.50
GPT-4o costs $3.50 more than Qwen3-Coder-480B-A35B-Instruct.
Benchmark Matrix
Reported model metrics from published sources.
Active Parameters
35B
Reported active parameter count at inference time.
Popular comparisons
30 total featuring Qwen3-Coder-480B-A35B-Instruct.
Qwen3-Coder-480B-A35B-Instruct is Alibaba/Qwen’s large coding-focused open-weight MoE model, tracked here with Alibaba Cloud Model Studio hosted pricing for software-building and agentic coding comparisons.
What this model is
Qwen describes Qwen3-Coder-480B-A35B-Instruct as the most powerful first variant in the Qwen3-Coder line: a 480B-parameter Mixture-of-Experts coding model with 35B active parameters. It is aimed at agentic coding, repo-scale work, tool use, and software engineering tasks rather than general chat alone.
This is the Qwen page to use when the question is specifically about coding-agent cost and capability, not general-purpose Qwen chat.
Pricing notes
Alibaba Cloud Model Studio lists tiered International pricing for qwen3-coder-480b-a35b-instruct. The 0–32K token tier is $1.50 per 1M input tokens and $7.50 per 1M output tokens. The 32K–128K tier is $2.70 input and $13.50 output. The 128K–200K tier is $4.50 input and $22.50 output.
The calculator on this page uses the 0–32K tier by default because it is the lowest standard tier and the cleanest baseline for comparison. Longer-context coding tasks can cost more, so the tier must be checked before estimating real repo-scale agent runs.
This is Alibaba Cloud Model Studio hosted pricing, not OpenRouter or another reseller route.
Benchmarks and specs
The Qwen3-Coder announcement describes the model as a 480B-parameter MoE model with 35B active parameters, native 256K context, and up to 1M context with extrapolation methods. The Hugging Face card also lists 480B total parameters and 35B activated parameters, and says the model supports only non-thinking mode.
The same Qwen3-Coder announcement frames the model around agentic coding and repo-scale tasks. This page does not list specific benchmark scores until they are tied to a named source, test setup, and comparison conditions.
Best fit
Qwen3-Coder-480B-A35B-Instruct is best for coding assistant cost comparisons, long-context coding-agent experiments, tool-use workflows, and Qwen vs DeepSeek vs Kimi vs Claude/GPT cost comparisons.
The practical warning is that the headline tier may not match a long repo-scale run. Bigger context windows move into higher Alibaba pricing tiers.