Qwen2.5-72B-Instruct API pricing
Updated
Among the 3 providers we checked (prices read 2026-11-14 to 2026-11-15), the lowest blended standard price for Qwen2.5-72B-Instruct is DeepInfra at USD 0.36 per 1M input tokens and USD 0.40 per 1M output tokens. The median across 3 standard prices is USD 0.37 / USD 0.40. Each price is read from the provider's own pricing page.
Does Qwen sell Qwen2.5-72B-Instruct directly?
Qwen publishes no price for Qwen2.5-72B-Instruct on its own API that we could find, so this page compares hosts only.
Qwen2.5-72B-Instruct model facts
Facts from Qwen's model card for Qwen/Qwen2.5-72B-Instruct and the Hugging Face model API, read 2026-11-16. BroadNet did not verify them, and hosts may serve the model at a different precision. BroadRouter is BroadNet's site.
| Fact | As stated | Source |
|---|---|---|
| Open weights | Yes. The weights are in a public Hugging Face repository (Qwen/Qwen2.5-72B-Instruct) with no access gate. | Hugging Face model API, checked 2026-11-16Quote“"private": false, "gated": false, 37 .safetensors files listed” |
| Licence | Custom licence: qwen | Qwen model card metadata, checked 2026-11-16Quote“license: other; license_name: qwen” |
| Parameters in the weight files (Hugging Face count, not the maker's figure) | About 72.7 billion (72,706,203,648) | Hugging Face model API, checked 2026-11-16Quote“"safetensors": { "total": 72706203648 }” |
| Context length | 131,072 tokens (generation 8,192 tokens) | Qwen model card, checked 2026-11-16Quote“Context Length: Full 131,072 tokens and generation 8192 tokens” |
| Input and output | Text in, text out (the card's task tag is text-generation) | Qwen model card metadata, checked 2026-11-16Quote“pipeline_tag: text-generation” |
| Tool calling | Not found on Qwen's model card when we read it | checked 2026-11-16 |
| First published on Hugging Face (not necessarily the release date) | 2024-09-16 | Hugging Face model API, checked 2026-11-16Quote“"createdAt": "2024-09-16T11:56:31.000Z"” |
Qwen2.5-72B-Instruct price per provider
USD per 1M tokens. Standard prices are ranked on a blend of 3 input tokens to 1 output token, cheapest first; the dearest standard price is 59% above the cheapest. Free rows, the maker's serving version and other service tiers follow, tagged and unranked.
| # | Provider | Input | Cached input | Output | Blended 3:1 | Context | Precision | Source |
|---|---|---|---|---|---|---|---|---|
| 1 | DeepInfra cheapestQwen/Qwen2.5-72B-InstructNotes and source quote“$0.36 in $0.40 out / 1M tokens” | USD 0.36 | not published | USD 0.40 | USD 0.37 | 32,768 | FP8 | DeepInfra page checked 2026-11-14 re-checked 2026-11-16 |
| 2 | FeatherlessQwen/Qwen2.5-72B-InstructNotes and source quoteDeveloper plan (per-token billing, from 50 USD a month in credits). No cached price published. “PRICING: INPUT $0.37 / OUTPUT $0.4” | USD 0.37 | not published | USD 0.40 | USD 0.3775 | 32,768 | FP8 | Featherless page checked 2026-11-15 re-checked 2026-11-16 |
| 3 | SiliconFlowNotes and source quoteContext shown as 33K. “Qwen2.5-72B-Instruct 33K $0.59 $0.59” | USD 0.59 | not published | USD 0.59 | USD 0.59 | 32,768 | not stated | SiliconFlow page checked 2026-11-14 re-checked 2026-11-16 |
| — | DeepInfra flex / spotQwen/Qwen2.5-72B-InstructNotes and source quoteFlex service tier (0.8× standard). “Flex (0.8×) $0.288 $0.32” | USD 0.288 | not published | USD 0.32 | USD 0.296 | 32,768 | FP8 | DeepInfra page checked 2026-11-14 re-checked 2026-11-16 |
- Ranked: standard-tier prices only, on (3 × input + output) ÷ 4. "Cached input: not published" means the provider's page shows no cache price; budget cached tokens at the input price.
- Other tiers (flex / spot): batch runs asynchronously within hours; flex and spot are best-effort and can be slow or refused at capacity; priority and fast cost more for reliability or speed. Shown, not ranked.
- Automated re-check 2026-11-16: a weekly job re-fetches each source page and looks for the recorded quote or prices. It is not a human review. "Re-checked" means it still found them. Rows with no re-check note show only the date the price was read.
- Precision is shown only where the provider states it; "not stated" is unknown, not full precision.
Report a wrong price or number (email to info@broadnet.ai, with this page's link filled in).
What does Qwen2.5-72B-Instruct cost for a real workload?
Monthly cost over 30 days at each standard price, using the presets of the LLM API cost calculator. Providers without a cached price bill cached input at the input price.
- Support chatbot: 2,000 requests a day, 1,500 input tokens (60% cached), 300 output.
- RAG over documents: 1,000 requests a day, 8,000 input tokens (30% cached), 500 output.
- Coding agent (one developer): 1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning.
| Provider | Support chatbot | RAG over documents | Coding agent (one developer) |
|---|---|---|---|
| DeepInfra | USD 39.60 | USD 92.40 | exceeds context |
| Featherless | USD 40.50 | USD 94.80 | exceeds context |
| SiliconFlow | USD 63.72 | USD 150.45 | exceeds context |
For the support chatbot, the cheapest price (DeepInfra, USD 39.60 a month) saves USD 0.9 against the median (USD 40.50). For the coding agent the gap is USD 18.00 a month (3% lower).
Which Qwen2.5-72B-Instruct provider is cheapest for your traffic?
DeepInfra has the lowest input price (USD 0.36) and the lowest output price (USD 0.40), so it is the cheapest standard price at every mix of input and output tokens, before cache discounts.
| Provider | No cache hits | 50% cached | 90% cached |
|---|---|---|---|
| DeepInfra (no cache price) | USD 0.37 cheapest | USD 0.37 cheapest | USD 0.37 cheapest |
| Featherless (no cache price) | USD 0.3775 | USD 0.3775 | USD 0.3775 |
| SiliconFlow (no cache price) | USD 0.59 | USD 0.59 | USD 0.59 |
USD per 1M tokens, 3 input tokens to 1 output token, each provider's cheapest standard row. A provider without a cached price bills cached input at the input price. Any charge for writing to the cache, which some providers add, is not included.
None of the 3 providers publishes a cached-input discount for Qwen2.5-72B-Instruct, so cache hits do not change the ranking.
All 3 providers offer 32,768 tokens of context.
Estimate your own bill
- LLM API cost calculator with Qwen2.5-72B-Instruct preloaded: your traffic and cache rate, against any of 428 models.
- Qwen token counter: count the tokens in your real prompts. Tokenizers differ, so the same text costs a different number of tokens on different models.
- Prompt caching calculator: what a higher cache-hit rate saves.
How does Qwen2.5-72B-Instruct compare with other Qwen models?
The Qwen models whose cheapest standard price is closest to Qwen2.5-72B-Instruct's (USD 0.37 blended):
| Model | Cheapest input | Cheapest output | Max context | Providers |
|---|---|---|---|---|
| Qwen3 VL 235B A22B | USD 0.20 | USD 0.88 | 262,144 | 3 |
| Qwen-MT Plus | USD 0.25 | USD 0.75 | 16,384 | 2 |
| Qwen3.5 35B A3B | USD 0.14 | USD 1 | 262,144 | 7 |
| Qwen3 Next 80B A3B Instruct | USD 0.09 | USD 1.10 | 262,144 | 5 |
| Qwen3 235B A22B FP8 | USD 0.17 | USD 1.09 | 32,768 | 2 |
| Qwen3.8 27B | USD 0.039 | USD 1.25 | 1,000,000 | 15 |
Priced closest from other makers: MiniMax M2.7 (MiniMax, USD 0.3675) · DeepSeek R1 Distill Llama 70B (DeepSeek, USD 0.375) · Llama 4 Maverick 17B 128E Instruct FP8 (Meta, USD 0.3875).
Compare Qwen2.5-72B-Instruct with
- Llama 3.3 70B vs Qwen2.5 72B: API price side by side
- Qwen2.5 72B vs Qwen 2.5 7B: API price side by side
Also compared: vs Llama 3.3 70B · vs Qwen 2.5 7B.
Every Qwen2.5-72B-Instruct price comparison · All model-vs-model price comparisons
All 428 models: API prices compared · Cheapest LLM APIs right now · LLM API price index (open data) · Download all prices (CSV, JSON) · Price change log
Questions
How much does the Qwen2.5-72B-Instruct API cost per token?
From USD 0.36 per 1M input tokens (USD 0.00000036 per token) and USD 0.40 per 1M output tokens (USD 0.0000004 per token) at the lowest standard prices we found. The median of 3 standard prices is USD 0.37 input and USD 0.40 output. Each price was read from the provider's own pricing page; the source is linked in its row.
Which provider is cheapest for Qwen2.5-72B-Instruct?
Among the 3 providers we checked (prices read 2026-11-14 to 2026-11-15), the lowest is DeepInfra, ranked on a blend of 3 input tokens to 1 output token (USD 0.37 per 1M). For input alone the lowest price is USD 0.36 (DeepInfra); for output alone, USD 0.40 (DeepInfra and 1 other, tied). Batch, flex, priority and fast tiers are shown in the table but not ranked.
Does Qwen sell Qwen2.5-72B-Instruct on its own API?
Qwen publishes no price for Qwen2.5-72B-Instruct on its own API that we could find, so there is no maker price to compare. The lowest-priced host we checked is DeepInfra at USD 0.36 / USD 0.40 per 1M tokens.
Does Qwen2.5-72B-Instruct have cached input pricing?
None of the 3 standard prices includes a cached-input price, so budget repeated prompts at the full input price.
What is the Qwen2.5-72B-Instruct context window?
32,768 tokens on every provider that states it.
How much does a coding agent cost per month on Qwen2.5-72B-Instruct?
For 1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning, over 30 days: USD 684.00 a month at DeepInfra, USD 702.00 at the median price and USD 1,115 at SiliconFlow. Try your own numbers.
Is there a batch, flex or priority price for Qwen2.5-72B-Instruct?
Yes: flex / spot from DeepInfra at USD 0.288 / USD 0.32 per 1M input / output tokens. These tiers trade speed or availability for price (batch runs within hours, flex and spot can be slow or refused at capacity, priority and fast cost more for speed or reliability), so they are shown but not ranked.
Is Qwen2.5-72B-Instruct free to use through an API?
No provider in our data lists a free price for Qwen2.5-72B-Instruct as of 2026-11-15. Free tiers and credits by provider (checked 2026-11-14) lists general offers that do not name a model; all free LLM API models are on one page.
Is Qwen2.5-72B-Instruct open source, with open weights, and what is its licence?
Qwen publishes the weights on Hugging Face as Qwen/Qwen2.5-72B-Instruct, in a public repository with no access gate. The licence, as the Qwen model card metadata states it: Custom licence: qwen. Open weights means you can download and run the model yourself; what you may do with it, including commercial use, is set by that licence, so read it in full (checked 2026-11-16). These facts come from Qwen's card and the Hugging Face model API; BroadNet did not verify them, and hosts may serve the model at a different precision.
How many parameters does Qwen2.5-72B-Instruct have?
Qwen's model card does not state a total that we found. Hugging Face counts about 72.7 billion (72,706,203,648) parameters in the published weight files of Qwen/Qwen2.5-72B-Instruct. That is a file-level count, not the maker's figure, and it can include extra modules (checked 2026-11-16).
How do I call Qwen2.5-72B-Instruct with the OpenAI SDK?
BroadNet does not serve Qwen2.5-72B-Instruct. Use the OpenAI-compatible base URL of a provider that lists it: create the client with OpenAI(base_url=..., api_key=...) and pass that provider's model ID. Documented base URLs for providers on this page, in alphabetical order: DeepInfra (https://api.deepinfra.com/v1/openai, model Qwen/Qwen2.5-72B-Instruct). Each provider page quotes the provider's docs with a check date.