Anthropic’s Claude and OpenAI’s GPT models, matched tier by tier, with each lab’s list price and what the same tokens cost on Zurelay, where both run on one API key.
By tier
Fable is Anthropic’s tier above Opus. Astra is the top of OpenAI’s GPT-6 family.
For the typical load: GPT-6 Astra costs 29% less than Claude Fable 5.1 ($112.50 against $157.50).
Anthropic built Opus 5.5 for long-running agentic coding. OpenAI says GPT-6.1 Sol nearly matches Astra on agentic coding.
For the typical load: GPT-6.1 Sol costs 64% less than Claude Opus 5.5 ($22.50 against $63.00).
Haiku 4.5 is Anthropic’s fastest model. Luna is the fastest, lowest-cost GPT-6.
For the typical load: GPT-6 Luna costs 95% less than Claude Haiku 4.5 ($1.13 against $24.30).
The typical load is 10,000 requests of 2,000 tokens in and 500 out. Struck-through prices are each lab’s list price.
Every model
| Model | Context | Input / 1M | Output / 1M | 10K requests | You save |
|---|---|---|---|---|---|
| GPT-6 LunaOpenAI · gpt-6-luna | 1.05M | $0.025 | $0.125 | $1.125 | 75%off |
| GPT-5.6 LunaOpenAI · gpt-5.6-luna | 1.05M | $0.08 | $0.49 | $4.05 | 59%off |
| GPT-6.1 SolOpenAI · gpt-6.1-sol | 1.05M | $0.50 | $2.50 | $22.50 | 75%off |
| GPT-6 SolOpenAI · gpt-6-sol | 1.05M | $0.50 | $2.50 | $22.50 | 75%off |
| Claude Haiku 4.5Anthropic · claude-haiku-4-5 | 200K | $0.54 | $2.70 | $24.30 | 46%off |
| GPT-5.6 TerraOpenAI · gpt-5.6-terra | 1.05M | $0.54 | $3.26 | $27.10 | 73%off |
| Claude Sonnet 5.5Anthropic · claude-sonnet-5-5 | 1M | $0.80 | $4.00 | $36.00 | 60%off |
| GPT-5.6 SolOpenAI · gpt-5.6-sol | 1.05M | $1.14 | $5.70 | $51.30 | 72%off |
| Claude Opus 5.5Anthropic · claude-opus-5-5 | 1M | $1.40 | $7.00 | $63.00 | 65%off |
| Claude Opus 5Anthropic · claude-opus-5 | 1M | $1.75 | $8.75 | $78.75 | 65%off |
| Claude Opus 4.8Anthropic · claude-opus-4-8 | 1M | $1.75 | $8.75 | $78.75 | 65%off |
| Claude Opus 4.6Anthropic · claude-opus-4-6 | 1M | $1.75 | $8.75 | $78.75 | 65%off |
| Claude Opus 4.7Anthropic · claude-opus-4-7 | 1M | $1.78 | $8.93 | $80.25 | 64%off |
| GPT-6 AstraOpenAI · gpt-6-astra | 1.05M | $2.50 | $12.50 | $112.50 | 75%off |
| Claude Fable 5.1Anthropic · claude-fable-5-1 | 1M | $3.50 | $17.50 | $157.50 | 65%off |
| Claude Fable 5Anthropic · claude-fable-5 | 1M | $3.50 | $17.50 | $157.50 | 65%off |
Prices in US dollars per 1M tokens, live from the catalog, with each lab's list price struck through. 10K requests is 10,000 requests of 2,000 tokens in and 500 out, at each price. You save is on output tokens.
Try both
Benchmarks don’t tell you which model handles your prompts better. Run the same requests through both and compare the answers, the tokens and the bill.
On Zurelay, every Claude and GPT model is on the same key and the same OpenAI-compatible endpoint, so switching is a change of model name. Claude also answers on Anthropic’s Messages API, for Claude Code and the Anthropic SDKs.
from openai import OpenAIclient = OpenAI(base_url="https://api.zurelay.com/v1", api_key="YOUR_ZURELAY_KEY")prompt = [{"role": "user", "content": "Refactor this function: ..."}]# Same key, same code: only the model name changesfor model in ["claude-opus-5-5", "gpt-6.1-sol"]: reply = client.chat.completions.create(model=model, messages=prompt) print(model, reply.usage.total_tokens, reply.choices[0].message.content[:200])It depends on the tier. For 10,000 requests of 2,000 tokens in and 500 out on Zurelay: each lab's most capable model, GPT-6 Astra costs 29% less than Claude Fable 5.1 ($112.50 against $157.50); built for agentic coding, GPT-6.1 Sol costs 64% less than Claude Opus 5.5 ($22.50 against $63.00); fast and low-cost, GPT-6 Luna costs 95% less than Claude Haiku 4.5 ($1.13 against $24.30). Both are 46–75% below their labs' list prices.
Yes. One Zurelay key calls every Claude and GPT model. With an OpenAI SDK, point it at https://api.zurelay.com/v1 and change only the model name to switch between them. Claude also answers on Anthropic's Messages API, for Claude Code and the Anthropic SDKs.
Both labs build their top models for agentic coding: Anthropic describes Claude Opus 5.5 as built for long-running agentic coding, and OpenAI says GPT-6.1 Sol nearly matches GPT-6 Astra on agentic coding and computer use. Results vary by task, so run your own through both. On Zurelay that's a change of model name.
Yes. Each model is served over several routes run by independent sellers, and every route is tested before it serves requests and every day after, with identity, dated-knowledge and reasoning checks that catch a different or older model sold under the name you asked for.
Create a key in seconds, point the SDK you already use at Zurelay, and every request costs up to 90% less from the first token.