Llama 3.2 3B vs Qwen3-14B API pricing
Updated
Llama 3.2 3B: USD 0.0509 input, USD 0.335 output per 1M tokens (Featherless). Qwen3-14B: USD 0.07 input, USD 0.28 output per 1M tokens (SiliconFlow). At 3 input tokens per output token, Qwen3-14B costs 0.5% more than Llama 3.2 3B (USD 0.1225 vs USD 0.1219 per 1M).
How do Llama 3.2 3B and Qwen3-14B differ on price?
- Cheapest input price: Llama 3.2 3B USD 0.0509 against USD 0.07 per 1M tokens, so Qwen3-14B costs 1.38× as much (38% more).
- Cheapest output price: Llama 3.2 3B USD 0.15 against USD 0.24 per 1M tokens, so Qwen3-14B costs 1.6× as much (60% more).
- Among the 201 indexed models, Llama 3.2 3B ranks 27th cheapest and Qwen3-14B 29th on the 3:1 blend.
- The cheaper model changes at 1.6 input tokens per output token and 2.9 input tokens per output token.
- Neither model has a standard price with a cached-input discount, so prompt caching does not lower either bill.
- Qwen3-14B states up to 131,072 tokens of context and Llama 3.2 3B 128,000, within 10% of each other.
- Llama 3.2 3B has 4 providers with a standard price, 1 more than Qwen3-14B; 1 provider sells both (Featherless).
- On one account, Featherless charges USD 0.1219 for Llama 3.2 3B (its cheapest price) and USD 0.15 for Qwen3-14B (1.22× its cheapest price) at 3:1, the lowest combined price among the providers that sell both.
- Shopping around matters more for Llama 3.2 3B: its dearest standard price is 2.2× its cheapest, against 1.22× for Qwen3-14B.
- For the one-developer coding agent preset, the gap is USD 351.00 a year in Llama 3.2 3B's favour (Llama 3.2 3B USD 121.95 a month, Qwen3-14B USD 151.20).
- Per character of input, the gap depends on the language: Llama 3.2 3B costs 9% less than Qwen3-14B on Chinese text but 59% less than Qwen3-14B on Hindi, because the two tokenizers split those languages differently.
Llama 3.2 3B and Qwen3-14B prices side by side
USD per 1M tokens, standard tier. "Cheapest" is the lowest price at 3 input tokens to 1 output token; each line can come from a different provider.
| Price | Llama 3.2 3B | Qwen3-14B |
|---|---|---|
| Model page | Llama 3.2 3B Instruct Meta · 4 providers | Qwen3-14B Qwen · 3 providers |
| Cheapest standard price, blended 3:1 | USD 0.1219 lower USD 0.0509 in / USD 0.335 out at Featherless Featherless page checked 2026-11-15 | USD 0.1225 USD 0.07 in / USD 0.28 out at SiliconFlow SiliconFlow page checked 2026-11-14 |
| Cheapest input | USD 0.0509 lower Featherless source | USD 0.07 SiliconFlow source |
| Cheapest cached input | No standard price lists a cache price; cached tokens bill at the input price. | No standard price lists a cache price; cached tokens bill at the input price. |
| Cheapest output | USD 0.15 lower Amazon Bedrock source | USD 0.24 DeepInfra source |
| Maker's own API | Meta publishes no paid standard price on its own API that we found. | Qwen publishes no paid standard price on its own API that we found. |
| Median standard price | USD 0.1005 / USD 0.335 input / output, 4 prices | USD 0.12 / USD 0.24 input / output, 3 prices |
| Context window | 128,000 tokens, largest stated; from 32,768 at some providers | 131,072 tokens, largest stated; from 32,768 at some providers |
| Providers with a standard price | 4 | 3 |
- Lower marks the lower price on a line; tie, equal prices. Other tiers and free rows are on the model pages.
Report a wrong price or number (email to info@broadnet.ai, with this page's link filled in).
What do Llama 3.2 3B and Qwen3-14B cost for a real workload?
30 days at each model's cheapest standard price for the workload (presets of the cost calculator; cached input at the cached price where listed).
| Workload | Llama 3.2 3B | Qwen3-14B | Difference |
|---|---|---|---|
| Support chatbot 2,000 requests a day, 1,500 input tokens (60% cached), 300 output | USD 10.61 Featherless | USD 11.34 SiliconFlow | USD 0.73 Llama 3.2 3B lower |
| RAG over documents 1,000 requests a day, 8,000 input tokens (30% cached), 500 output | USD 17.24 Featherless | USD 21.00 SiliconFlow | USD 3.76 Llama 3.2 3B lower |
| Coding agent (one developer) 1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning | USD 121.95 Cloudflare Workers AI | USD 151.20 SiliconFlow | USD 29.25 Llama 3.2 3B lower |
Llama 3.2 3B is cheaper on all three workloads. The widest gap is on Coding agent (one developer): the dearer model costs 24% more than the cheaper one.
Does the cheaper model change with your input:output ratio?
The cheaper model changes with the mix: below 1.6 input tokens per output token, Llama 3.2 3B is cheaper; from 1.6 to 2.9, Qwen3-14B is cheaper; above 2.9, Llama 3.2 3B is cheaper. At the site's 3:1 blend, Llama 3.2 3B is cheaper. More input per output token, as with retrieval or agents that resend long contexts, favours the lower input price; long answers favour the lower output price.
| Input : output tokens | Llama 3.2 3B | Qwen3-14B |
|---|---|---|
| 1 : 3 (long answers) | USD 0.15 lower Amazon Bedrock | USD 0.21 DeepInfra |
| 1 : 1 | USD 0.15 lower Amazon Bedrock | USD 0.175 SiliconFlow |
| 3 : 1 (the site’s blend) | USD 0.1219 lower Featherless | USD 0.1225 SiliconFlow |
| 10 : 1 (retrieval, long prompts) | USD 0.07673 lower Featherless | USD 0.08909 SiliconFlow |
| 30 : 1 (agents, big contexts) | USD 0.06006 lower Featherless | USD 0.07677 SiliconFlow |
USD per 1M tokens at the cheapest standard provider for each mix, before cache discounts. Thinking tokens bill as output.
What do Llama 3.2 3B and Qwen3-14B cost per character in other languages?
Cost to send 1M characters at each model's cheapest non-promotional standard input price, with the Llama 3.3 70B tokenizer for Llama 3.2 3B and the Qwen3 32B tokenizer for Qwen3-14B:
| Content | Llama 3.2 3B | Qwen3-14B |
|---|---|---|
| English | USD 0.011 lower Featherless | USD 0.015 SiliconFlow |
| Arabic MSA | USD 0.020 lower Featherless | USD 0.027 SiliconFlow |
| Egyptian Arabic | USD 0.020 lower Featherless | USD 0.028 SiliconFlow |
| Hindi | USD 0.027 lower Featherless | USD 0.065 SiliconFlow |
| Chinese | USD 0.041 lower Featherless | USD 0.045 SiliconFlow |
| Spanish | USD 0.014 lower Featherless | USD 0.019 SiliconFlow |
| Code | USD 0.013 lower Featherless | USD 0.018 SiliconFlow |
From the real cost by language dataset (measured tokens per character on FLORES-200 and other texts). Your own text will differ.
Check it on your own traffic
- Llama 3.2 3B Instruct API pricing: all 4 providers and other tiers.
- Qwen3-14B API pricing: all 3 providers and other tiers.
- Cost calculator with Llama 3.2 3B and Qwen3-14B preloaded
- Llama 3 token counter
- Qwen3 token counter
- Prompt caching calculator
More comparisons: Llama 3.2 3B vs Llama 3.2 1B · Gemma 4 26B A4B vs Qwen3-14B · Qwen3 30B A3B vs Qwen3-14B · Qwen3-14B vs Qwen3-8B · every Llama 3.2 3B comparison · every Qwen3-14B comparison · all 662 price comparisons
Questions
Is Llama 3.2 3B cheaper than Qwen3-14B?
Yes. On a blend of 3 input tokens to 1 output token, Llama 3.2 3B costs USD 0.1219 per 1M tokens and Qwen3-14B USD 0.1225, at each model's cheapest standard price. For input alone, Llama 3.2 3B is cheaper (USD 0.0509 against USD 0.07 per 1M); for output alone, Llama 3.2 3B is (USD 0.15 against USD 0.24). This compares price only, not quality.
What do Llama 3.2 3B and Qwen3-14B cost per 1M tokens?
Llama 3.2 3B: from USD 0.0509 input and USD 0.335 output at Featherless; median of 4 standard prices USD 0.1005 / USD 0.335. Qwen3-14B: from USD 0.07 input and USD 0.28 output at SiliconFlow; median of 3 standard prices USD 0.12 / USD 0.24. Each price is read from the provider's own pricing page; the source is linked in the table.
Which providers sell both Llama 3.2 3B and Qwen3-14B?
1 provider has a standard price for both: Featherless (USD 0.1219 and USD 0.15 at 3:1). Buying each model from its own cheapest provider costs USD 0.1219 and USD 0.1225.
Which is cheaper for a coding agent or a chatbot, Llama 3.2 3B or Qwen3-14B?
For a coding agent (1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning): USD 121.95 a month on Llama 3.2 3B (Cloudflare Workers AI) against USD 151.20 on Qwen3-14B (SiliconFlow), a difference of USD 29.25. For a support chatbot (2,000 requests a day, 1,500 input tokens (60% cached), 300 output): USD 10.61 a month on Llama 3.2 3B (Featherless) against USD 11.34 on Qwen3-14B (SiliconFlow), a difference of USD 0.73. Each uses the cheapest standard price for that workload, cached input included. Try your own numbers.
At what input:output ratio does the cheaper of Llama 3.2 3B and Qwen3-14B change?
The cheaper model changes with the mix: below 1.6 input tokens per output token, Llama 3.2 3B is cheaper; from 1.6 to 2.9, Qwen3-14B is cheaper; above 2.9, Llama 3.2 3B is cheaper. At the site's 3:1 blend, Llama 3.2 3B is cheaper. More input per output token, as with retrieval or agents that resend long contexts, favours the lower input price; long answers favour the lower output price.
Do Llama 3.2 3B and Qwen3-14B have cached input pricing?
Llama 3.2 3B: no standard price lists a cached-input discount. Qwen3-14B: no standard price lists a cached-input discount. Cached tokens are billed at the input price where no cache price is published.
Which has the larger context window, Llama 3.2 3B or Qwen3-14B?
Qwen3-14B: up to 131,072 tokens, against 128,000 for Llama 3.2 3B, as stated on providers' pricing pages. Some providers offer less than the maximum; check the model pages before you pick the cheapest row for long prompts.