Qwen vs Claude: EU pricing, regions 2026

TL;DR

Qwen vs Claude in the EU is residency first. Claude Sonnet 5.5 costs 2 / 10 USD per million tokens plus 10 percent on regional endpoints and stays behind eu.anthropic.claude-sonnet-5-5. Qwen3 has no geo profile. Next 80B in Ireland is 0.18 / 1.41 USD. Coder Next in London is 0.78 / 1.86. Google MaaS for Qwen3 235B is global or US.

What separates Qwen and Claude for a company?

Qwen vs Claude in 2026 is two product lines with opposite residency models. Claude Sonnet 5.5 gives 1 million tokens of context behind an EU geo profile on Bedrock. Qwen3 gives Apache 2.0 weights and stays in the single Region you call. None of the Qwen cards we read publish a geo profile. The bare model ID is the residency control.

The Sonnet 5.5 card dated 28 September 2026 lists a 1 million token context, 128,000 output tokens and a June 2026 cutoff. The EU profile is eu.anthropic.claude-sonnet-5-5. Model EOL is no sooner than 28 September 2027. Which Claude model to start with is in Claude Sonnet vs Opus.

The Qwen3 32B card lists launch on 29 April 2025, EOL no sooner than 29 April 2026, 32,000 context tokens and 8,000 output tokens. The runtime ID is qwen.qwen3-32b-v1:0. The 235B card lists launch on 28 April 2025, 256,000 context tokens, 8,000 output tokens and ID qwen.qwen3-235b-a22b-2507-v1:0. Both cards still say Active. “EOL no sooner than” April 2026 is a floor that has already passed, not a published shutdown date.

Trait	Claude Sonnet 5.5	Qwen3 32B	Qwen3 235B A22B	Qwen3 Next 80B	Qwen3 Coder Next
Role	Production, agents	Dense general model	Frontier MoE	Fast MoE	Software engineering
Launch	28 September 2026	29 April 2025	28 April 2025	11 September 2025	4 February 2026
Context / output	1M / 128K	32K / 8K	256K / 8K	256K / 8K	256K / 16K
Bedrock ID	eu.anthropic.claude-sonnet-5-5	qwen.qwen3-32b-v1:0	qwen.qwen3-235b-a22b-2507-v1:0	qwen.qwen3-next-80b-a3b	qwen.qwen3-coder-next
EU on Bedrock	Geo profile	In-region, several EU Regions	In-region Frankfurt, Stockholm, Milan, London	In-region Milan, Ireland, London	In-region London
Weights	Closed	Apache 2.0	Apache 2.0	Closed on Bedrock	Closed on Bedrock

The Hugging Face 32B card lists 32.8 billion parameters, 32,768 tokens native and YaRN up to 131,072. The 235B card lists 235 billion parameters total, 22 billion active, 128 experts, 8 active and 262,144 tokens native. Both licences are Apache 2.0, Copyright 2024 Alibaba Cloud. The Next 80B card lists 80 billion parameters total and 3 billion active, launch 11 September 2025. The Coder Next card lists launch 4 February 2026 and EOL no sooner than 4 February 2027. For a 2027 plan, Coder Next is the Qwen row with a documented horizon. The other open-weight comparison, with no EU region for the frontier model, is Mistral vs Claude.

What does Qwen vs Claude cost in the EU?

Qwen vs Claude on Bedrock is an order-of-magnitude price gap. Anthropic lists Sonnet 5.5 at 2 / 10 USD per million tokens, regional endpoints at 10 percent more, so 2.20 / 11 USD in the EU. Next 80B in Ireland and Milan is 0.18 / 1.41 USD. That is the documented EU rate for a general-purpose MoE.

On-demand standard prices on 14 October 2026, USD per million tokens:

Route	Input / output	Cache read	Note
Claude Sonnet 5.5, Claude API global	2.00 / 10.00	0.10	5-minute write 2.50, 1-hour write 4.00
Claude Sonnet 5.5, Bedrock EU or Google Europe	2.20 / 11.00	0.11 (10 percent rule)	Batch is 50 percent of the EU row
Claude Opus 5.5, Claude API global	4.00 / 20.00	0.20	regional endpoints plus 10 percent
Qwen3 Next 80B, US East/Ohio/West	0.15 / 1.20	not read	no geo profile
Qwen3 Next 80B, Ireland and Milan	0.18 / 1.41	not read	compatibility lists these two plus London
Qwen3 Next 80B, London	0.23 / 1.86	not read	highest EU Next rate we read
Qwen3 Coder Next, US East/Ohio/West	0.50 / 1.20	not read	compatibility: N. Virginia, London, Sydney
Qwen3 Coder Next, Ireland and Milan	0.60 / 1.44	not read	price row; compatibility lists London
Qwen3 Coder Next, London	0.78 / 1.86	not read	only EU row on the compatibility page
Qwen3 Coder Next, Frankfurt	0.60 / 1.44	not read	pricing page only, not the compatibility list
Qwen3 32B, Sydney	0.1545 / 0.6180	not read	only 32B on-demand row that page showed
Qwen3 235B, Sydney	0.2266 / 0.9064	not read	only 235B on-demand row that page showed

The Bedrock pricing page sets Priority at a 75 percent premium and Flex and Batch at a 50 percent discount to Standard. For 32B and 235B the region tab in this run only rendered Sydney. We do not invent an EU dollar rate for those two IDs. The full Claude matrix is in Claude API pricing in the EU.

The ticket rows below are our arithmetic, not a vendor quote. Assumption: 3,000 input and 400 output tokens, 10,000 jobs a month.

Workload	Sonnet 5.5 global	Sonnet 5.5 EU	Next 80B Ireland	Coder Next London
10,000 support replies	100 USD	110 USD	11 USD	31 USD
1,000 reviews at 40,000 / 2,000 tokens	100 USD	110 USD	10 USD	35 USD

Next 80B wins on price when Milan or Ireland as a single Region is enough. Coder Next in London stays cheaper than Sonnet and costs two to three times Next. Claude wins when you need 1 million tokens of context, a geo profile across several EU Regions, and an EOL no sooner than September 2027.

Which EU regions keep Qwen, which keep Claude?

Qwen stays in the Region you set on the client. Claude stays in a group of European Regions behind the eu. profile. The compatibility page lists in-region for every Qwen model and marks geo and global as unsupported. From Frankfurt, Claude may land in Stockholm, Milan, Spain, Ireland or Paris. Qwen3 32B in Frankfurt stays in Frankfurt.

Compatibility page on 14 October 2026, in-region only, no geo column:

Model	EU Regions on the compatibility page
Qwen3 32B	Frankfurt, Stockholm, Milan, Ireland, London
Qwen3 235B A22B 2507	Frankfurt, Stockholm, Milan, London
Qwen3 Next 80B A3B	Milan, Ireland, London
Qwen3 VL 235B A22B	Milan, Ireland, London
Qwen3 Coder Next	London
Qwen3 Coder 30B A3B	Frankfurt, Stockholm, Milan, Ireland, London
Qwen3 Coder 480B A35B	Stockholm, London
Claude Sonnet 5.5	EU geo from Frankfurt, Zurich, Stockholm, Milan, Spain, Ireland, London, Paris

Next 80B has no Frankfurt row on the compatibility page. Pricing agrees: the Frankfurt table for that family only names Coder Next. If Frankfurt is the only allowed Region, take 32B, 235B or Coder 30B, not Next 80B. Coder Next has only London as an EU row on compatibility. Pricing also publishes Frankfurt, Ireland and Milan. We treat compatibility as the residency source and pricing as the rate source.

Google MaaS is not an EU route for Qwen3 235B. The MaaS card lists qwen3-235b-a22b-instruct-2507-maas in global and us-south1, with processing in the US multi-region. The endpoint has been deprecated since 21 July 2026 and retires on 21 October 2026. Claude on Google Europe stays the residency route. Qwen on Google does not, in this run.

Foundry lists Qwen3-32B with Data Zone and Global, not Standard regional. We did not read a public dollar rate or a full EU region list on the Foundry models page. How Bedrock geo profiles work is in the Amazon Bedrock EU guide. The contract path is in the Claude GDPR comparison.

How do you call Qwen and Claude on Bedrock?

Call Qwen with the bare model ID in an EU Region and Claude with the eu. profile. There is no geo prefix for Qwen. qwen.qwen3-32b-v1:0 from eu-central-1 stays in Frankfurt. eu.anthropic.claude-sonnet-5-5 from Frankfurt may land in six EU Regions. Both clients are the same bedrock-runtime.

Five steps for a first comparison run in Frankfurt:

1. Create the runtime client in eu-central-1 and deny global. profiles in IAM.
2. Enable model access for qwen.qwen3-32b-v1:0 and the inference profile eu.anthropic.claude-sonnet-5-5.
3. Set max_tokens explicitly. Qwen3 32B allows 8,000 output tokens, Sonnet 5.5 allows 128,000.
4. Log modelId and the AWS Region next to the request ID.
5. Run ten identical tickets on both IDs before you change the default.

import boto3

client = boto3.client("bedrock-runtime", region_name="eu-central-1")

qwen = client.converse(
    modelId="qwen.qwen3-32b-v1:0",
    messages=[{"role": "user", "content": [{"text": "Summarise Bedrock EU residency in two sentences."}]}],
    inferenceConfig={"maxTokens": 512},
)

claude = client.converse(
    modelId="eu.anthropic.claude-sonnet-5-5",
    messages=[{"role": "user", "content": [{"text": "Summarise Bedrock EU residency in two sentences."}]}],
    inferenceConfig={"maxTokens": 512},
)

for label, response in (("qwen", qwen), ("claude", claude)):
    for block in response["output"]["message"]["content"]:
        if "text" in block:
            print(label, block["text"])
    print(label, response["usage"])

Next 80B does not belong in this Frankfurt client. Compatibility names Milan, Ireland and London. Coder Next belongs in London if you follow the compatibility list. A gateway that maps internal names onto these IDs is in the LiteLLM EU guide.

When to self-host Qwen, when to keep Claude on the EU profile?

Self-host when the weights must sit in your data centre. Keep Claude on the EU profile when personal prompts must stay in the EU and you will not run GPUs. Use Qwen on Bedrock in-region when one AWS data centre is enough. The three paths can run together.

Hugging Face documents vllm serve Qwen/Qwen3-32B for 32B and vllm serve Qwen/Qwen3-235B-A22B-Instruct-2507 --tensor-parallel-size 8 --max-model-len 262144 for 235B. A 1 million token context on 235B needs about 1000 GB of GPU memory on that card. That is operations, not an API call. The GPU path for another open model is in DeepSeek V4.1 self-hosted. Token prices without your own hardware are in DeepSeek pricing on Bedrock EU.

We would take Claude Sonnet 5.5 as the default for personal prompts once the DPO requires EU processing and a single data centre is too tight. Qwen3 32B in Frankfurt is the cheap classifier on the same account. Next 80B in Ireland or Milan is the cheap generator when Frankfurt is not mandatory. Coder Next in London is the Qwen code row with EOL no sooner than February 2027. We would not plan Qwen3 235B on Google MaaS after 21 October 2026.

FAQ

What does Qwen vs Claude cost for 10,000 support tickets?

With 3,000 input and 400 output tokens per ticket, Sonnet 5.5 costs 100 USD globally and 110 USD in the EU. Next 80B in Ireland costs 11 USD. Coder Next in London costs 31 USD. Those ticket rows are our arithmetic from the on-demand rates on 14 October 2026, not a vendor quote.

Qwen vs Claude: which model should a European company take first?

Claude Sonnet 5.5 behind eu.anthropic.claude-sonnet-5-5 when personal prompts must stay in the EU and a single data centre is too tight. Qwen3 32B in Frankfurt when classification in one Region is enough. Next 80B in Ireland or Milan when input must stay under 0.20 USD and Frankfurt is not mandatory.

Does Qwen3 stay in the EU on Amazon Bedrock?

Yes, as an in-region call, not as a geo profile. 32B lists Frankfurt, Stockholm, Milan, Ireland and London. 235B lists Frankfurt, Stockholm, Milan and London. Next 80B lists Milan, Ireland and London, not Frankfurt. Coder Next lists only London on compatibility. The cards do not publish an eu.qwen. profile.

Self-host Qwen3 235B or run it on Bedrock?

Self-host when the weights must sit in your data centre. The Hugging Face card names Apache 2.0, 235 billion parameters, 22 billion active and vllm with --tensor-parallel-size 8. Bedrock in-region when you will not run GPUs and Frankfurt, Stockholm, Milan or London is enough. Google MaaS disappears on 21 October 2026 and never had an EU region.

Qwen on Google or Claude on Google?

Claude Sonnet 5.5 has a Europe multi-region at a 10 percent premium. Qwen3 235B MaaS sits in global and us-south1, has been deprecated since 21 July 2026 and retires on 21 October 2026. The same frontier comparison on Google in the EU has no Qwen row.

Does Qwen3 32B have an EU price on Bedrock?

On 14 October 2026 the pricing page only showed the Sydney row for 32B and 235B (0.1545 / 0.6180 and 0.2266 / 0.9064). Compatibility lists both models in Frankfurt. We leave the EU dollar rate blank until AWS publishes an EU row.

Sources

1. Anthropic: Pricing (14 October 2026)
2. AWS: Claude Sonnet 5.5 model card (14 October 2026)
3. AWS: Qwen3 32B model card (14 October 2026)
4. AWS: Qwen3 235B A22B 2507 model card (14 October 2026)
5. AWS: Qwen3 Next 80B A3B model card (14 October 2026)
6. AWS: Qwen3 Coder Next model card (14 October 2026)
7. AWS: Regional availability by models (14 October 2026)
8. Amazon Bedrock Pricing (14 October 2026)
9. Hugging Face: Qwen3-32B (14 October 2026)
10. Hugging Face: Qwen3-235B-A22B-Instruct-2507 (14 October 2026)
11. Google Cloud: Qwen3 235B MaaS (14 October 2026)
12. Microsoft Learn: Foundry models sold by Azure (14 October 2026)